阅读说明：本技术 具有可变施用率的种植物田地的种植物处理方法 (Method for treating plants in a field of plants with variable application rates ) 是由 M·坦普尔 于 2020-03-27 设计创作，主要内容包括：一种用于种植物田地的种植物处理的方法,所述方法包括：基于与所述种植物田地(300)上的预期条件有关的离线田地数据(Doff)确定施用率决策逻辑(10)；拍摄(S20)种植物田地(300)的种植物的图像(20)；识别(S30)所拍摄的图像(20)上的对象(30)；基于所确定的施用率决策逻辑(10)和所识别的对象(30)确定施用率；以及基于所确定的施用率确定(S40)用于控制处理装置(200)的处理布置(270)的控制信号(S)。(A method for plant treatment of a field of plants, the method comprising: determining application rate decision logic (10) based on offline field data (Doff) relating to expected conditions on said field of plants (300); capturing (S20) an image (20) of a plant of a field (300) of plants; identifying (S30) an object (30) on the captured image (20); determining an application rate based on the determined application rate decision logic (10) and the identified object (30); and determining (S40) a control signal (S) for controlling a processing arrangement (270) of the processing device (200) based on the determined application rate.)

1. A method for plant treatment of a field of plants, the method comprising:

determining application rate decision logic (10) based on offline field data (Doff) relating to expected conditions on said field of plants (300);

capturing (S20) an image (20) of a plant of a field (300) of plants;

identifying (S30) an object (30) on the captured image (20);

determining an application rate based on the determined application rate decision logic (10) and the identified object (30); and

determining (S40) a control signal (S) for controlling a processing arrangement (270) of the processing device (200) based on the determined application rate.

2. The method of claim 1, wherein,

capturing (S20) an image (20) of a plant of a field (300) of plants; identifying (S30) an object (30) on the captured image (20); determining (S40) an application rate and determining (S50) control signals (S) for controlling a processing arrangement (270) are performed as a real-time process such that the processing device (200) is instantaneously controllable based on captured images of the field of plants as the processing device traverses the field (300) while processing at specific locations of the field.

3. The method of any one of the preceding claims,

the application rate decision logic (10) provides logic for determining an application rate for treating the plant in dependence on an expected loss of efficacy of a crop to be planted in the field of plants.

4. The method according to any one of claims 1 to 3, wherein the application rate decision logic (10) comprises a variable application rate dependent on one or more parameters derived from the image and/or object recognition.

5. The method of any one of claims 1 to 4,

the recognition object (30) models the geometry of the object on the basis of the captured image (10) and/or the geometrical object contour, on the basis of its type and/or its growth phase.

6. The method of any one of the preceding claims,

determining the application rate based on the identified object includes determining an object type, an object growth stage, and/or an object density.

7. The method according to any one of the preceding claims, comprising:

receiving, by said processing device (200), online field data (Don) relating to current conditions on said field of plants (300); and

determining the control signal (S) depending on the determined application rate decision logic (10) and the determined recognition object (30) and/or the determined online field data (Don).

8. The method of claim 7, wherein,

the online field data (Don) relates to current weather condition data, current plant growth data and/or current soil data.

9. The method according to any one of the preceding claims, comprising the steps of:

providing verification data (V) depending on a performance review of the treatment of the plant; and

adjusting the application rate decision logic (10) in dependence on the verification data (V).

10. The method of claim 9, comprising the steps of:

adjusting the geometric object contour based on the verification data (V).

11. The method according to any one of the preceding claims, comprising the steps of:

adjusting the application rate decision logic (10) using a machine learning algorithm.

12. A field manager system (100) for a processing device (200) for plant processing of a field (300) of plants, comprising:

an offline field data interface (150) adapted to receive offline field data (Doff) relating to expected conditions on said field of plants (300);

a machine learning unit (110) adapted to determine an application rate decision logic (10) of the processing device (200) depending on the offline field data (Doff); and

a decision logic interface (140) adapted to provide the application rate decision logic (10) to a processing device (200) according to any of claims 14 to 18.

13. The field manager system of claim 12, comprising:

an authentication data interface (160) adapted to receive authentication data (V), wherein,

the machine learning unit (110) is adapted to adjust the application rate decision logic (10) in dependence of the verification data (V).

14. A treatment device (200) for plant treatment of plants, comprising:

an image capturing device (220) adapted to take an image (20) of a plant;

a decision logic interface (240) adapted to receive application rate decision logic (10) from the field manager system (100) of claim 10;

a processing arrangement (270) adapted to process the plant in dependence of the received application rate decision logic (10);

an image recognition unit (230) adapted to recognize an object (30) on the captured image (20);

a process control unit (210) adapted to determine a control signal (S) for controlling the processing arrangement (270) in dependence on the received application rate decision logic (10) and the identified object (30);

wherein the decision logic interface (240) of the processing device (200) is connectable to a decision logic interface (140) of a field manager system (100) according to claim 10; and

wherein the processing device (200) is adapted to activate the processing arrangement (270) based on the control signal (S) of the processing control unit (210).

15. The process control device of claim 14, comprising:

an online field data interface (240) adapted to receive online field data (Don) related to current conditions on said field of plants (300), wherein,

-the process control unit (210) is adapted to determine a control signal (S) for controlling a processing arrangement (270) depending on the received application rate decision logic (10) and the identified object (30) and/or the online field data (Don);

16. the processing apparatus according to any one of claims 14 or 15,

wherein the image capturing device (220) comprises one or more cameras, in particular on a boom of the processing device (200), wherein the image recognition unit (230) is adapted to recognize objects using red-green-blue RGB data and/or near infrared NIR data.

17. The processing apparatus according to any one of claims 14 to 16,

wherein the treatment device (200) is designed as a smart nebulizer, wherein the treatment arrangement (270) is a nozzle arrangement.

18. The processing apparatus according to any one of claims 14 to 17,

wherein the image capturing device (220) comprises a plurality of cameras and the processing arrangement (270) comprises a plurality of nozzle arrangements, each nozzle arrangement being associated with one of the plurality of cameras such that an image captured by the camera is associated with an area to be processed by the respective nozzle arrangement.

19. A processing system comprising a field manager system according to any of claims 12 or 13 and a processing device according to any of claims 14 to 18.

Technical Field

The present invention relates to a method and a processing device for plant processing of a field of plants, and a field manager system for such a processing device and processing system.

Background

The general background of the invention is the treatment of plants in agricultural fields. The treatment of plants, in particular of actual crops, also includes the treatment of weeds in agricultural fields, the treatment of insects in agricultural fields and the treatment of pathogens in agricultural fields.

Agricultural machines or automated treatment devices, like intelligent sprayers, treat weeds, insects, and/or pathogens in agricultural fields based on ecological and economic regulations. In order to automatically detect and identify different objects to be processed, image recognition is used.

The smart sprayer has a camera sensor attached to detect green plants by distinguishing the creature from the soil. This enables them to apply the herbicide only where the organisms (i.e. weeds) are detected. To this end, the intelligent sprayer turns off and closes a single nozzle as it moves through the field.

The biological detection technique can provide a percent biological coverage in each nozzle pass, e.g., 2% is biological covered in 1 square meter measured in front of the nozzle. Furthermore, it can provide geometrical information about each individual detection object, for example, size, shape or pattern, which sums up the total percentage values.

The nozzle of the intelligent sprayer is turned on and off only where application is desired, and significant savings in crop protection products can be achieved. To achieve high weed control efficacy, the application rate of the herbicide or tank mix of herbicides must be high enough to be able to control the most difficult weeds present in the field. Different application rates are required for each weed species, weed growth stage, weed density, and different environmental conditions to achieve the agronomically sustainable target efficacy.

Current sprayers work at a single application rate that has been predefined when creating a tank mix of water and herbicide.

Disclosure of Invention

It would be advantageous to have an improved method of plant treatment for a field of plants having a variable application rate.

The object of the invention is solved with the subject matter of the independent claims, wherein further embodiments are comprised in the dependent claims. It should be noted that the following described aspects and examples of the invention also apply to the method, the processing device and the field manager system.

According to a first aspect, a method for plant treatment of a field of plants comprises:

receiving application rate decision logic based on offline field data relating to expected conditions on a field of plants;

taking a plant image of a plant field;

identifying an object on the captured image;

determining an application rate based on the determined application rate decision logic and the identified subject;

determining a control signal for controlling a processing arrangement of the processing device based on the determined application rate;

as used herein, plant treatment preferably comprises: protection crops, which are cultivated plants on a field of plants; the elimination of weeds which are uncultivated and may be harmful to the crop, in particular with herbicides; killing insects on crops and/or weeds, particularly with insecticides; and to destroy any pathogens on the crops and/or weeds, like diseases, in particular with fungicides; and regulating the growth of plants, particularly by using plant growth regulators. As used herein, the term "insecticide" also encompasses nematicides, acaricides, and molluscicides. In addition, safeners can be used in combination with herbicides.

In one embodiment, capturing the image includes capturing an image associated with a particular location on or in the field of the plant to be treated in real time. In this way, fine-tuning of different conditions across the field can be performed in near real-time as the process is performed. Furthermore, the treatments can be applied in a very targeted manner, which results in a more efficient and sustainable agriculture. In a preferred embodiment, the processing device comprises a plurality of image capturing devices configured to capture images of the field of plants as the processing device traverses the field of plants. Each image captured in this manner may be associated with a location and thus provide a snapshot of the real-time situation in the location of the field of plants to be processed. To enable real-time, location-specific control of a processing device, parameter packets received prior to processing provide a way to accelerate situation-specific control of the processing device. Thus, decisions can be made on-the-fly as the processing device traverses the field and captures location-specific images of the field location to be processed.

Preferably, the steps of capturing images, determining control signals and optionally providing control signals to the control unit to initiate the process are performed in real time during passage of the processing means through the field or during processing of the field. Alternatively, control signals may be provided to the control unit of the processing device to initiate the processing of the field of plants.

In one embodiment, offline data is received that includes expected environmental data, such as expected soil data, time predictions as to when to spray based on expected weather data, and/or agricultural action specifications, e.g., the farming system in use; the expected spectrum of weeds; and the expected weed growth stage. Determining the application rate can also be based on offline data, including expected environmental data, expected weed spectrum, and expected weed growth stage.

As used herein, the term "subject" includes plants, like weeds or crops, insects and/or pathogens. The object may refer to an object to be treated by the treatment device, such as a plant, like a weed or a crop, an insect and/or a pathogen. The object may be treated with a treatment product, such as a crop protection product. The objects may be associated with locations in the field to allow location-specific processing.

Preferably, control signals for controlling the processing means may be determined based on the determined application rate decision logic, the identified objects and the online field data. In one embodiment, the online field data is collected in real time, particularly by the plant processing device. Collecting online field data may include collecting sensor data from sensors attached to the processing device or placed in a field of plants, particularly on-the-fly or in real-time as the processing device passes through the field. Collecting online field data may include the growth stage of weeds relative to surrounding crops; seed propagation; soil data collected via soil sensors in the field associated with soil characteristics, such as current soil conditions, e.g., nutrient content, soil moisture, and/or soil composition; or weather data collected via weather sensors placed in or near the field or attached to the processing device and associated with current weather conditions or data collected via soil and weather sensors.

Seed reproduction involves an indication of the location where weeds have reached seed maturity in the past, with the assumption that a higher density of weeds is expected at this location due to the seeds of the weeds remaining on the ground. Furthermore, it is hypothesized that weed seeds may have spread as a result of working in the field of plants. This information may be collected in a ground weed seed database. Thus, the application rate decision logic increases the application rate each time seed reproduction is determined at a location.

The term "offline field data" as used herein refers to any data generated, collected, processed prior to determining application rate decision logic. Preferably, the offline field data is determined externally from the plant processing device, or prior to use of the processing device. For example, the offline field data includes weather data relating to expected weather conditions at the time of processing; expected soil data relating to expected soil conditions at the time of treatment, such as nutrient content or soil moisture, and/or soil composition; growth stage data relating to the growth stage of, for example, weeds or crops at the time of treatment; disease data relating to disease stage of the crop at the time of treatment; the level of resistance of the weeds and/or the yield impact of the weeds on the crop.

The term "spatially resolved" as used herein refers to any information on the sub-field scale. Such resolution may be associated with more than one location coordinate on a plant field or with a spatial grid of a plant field having grid elements on a sub-field scale. In particular, information on a field of plants may be associated with more than one location or grid element on the field of plants. Such spatial resolution on the sub-field scale allows for more customized and targeted processing of the field of plants.

The term "conditions on a plant field" refers to any condition of the plant field or environmental conditions in the plant field that has an effect on the treatment of the plant. Such conditions may be associated with soil or weather conditions. Soil conditions may be specified by soil data relating to current or expected conditions of the soil. The weather conditions may be associated with weather data relating to current or expected conditions of the weather. The growth conditions may be associated with, for example, the growth stage of the crop or weed. The disease condition can be correlated with disease data relating to the current or expected condition of the disease.

The term "treatment device" as used herein or also referred to as control technology preferably includes chemical control technology. The chemical control technique preferably comprises at least one device for applying a treatment product, in particular a crop protection product like a pesticide and/or herbicide and/or plant growth regulator and/or fungicide. Such apparatus may include a treatment arrangement of one or more spray guns or nozzles arranged on agricultural machines, drones or robots for maneuvering through a field of plants. In a preferred embodiment, the processing device comprises one or more spray guns and associated image capture device(s). The image capture device may be arranged such that the image is associated with an area to be processed by the one or more spray guns. The image capturing device may for example be mounted such that an image in the direction of travel of the processing device is taken, covering the area to be processed by the respective nozzle(s). Each image may be associated with a location and as such provides a snapshot of the real-time condition of the field of plants prior to processing. Thus, the image capturing device may take an image of a specific location of the field of plants as the processing device traverses the field, and the control signal may be adapted accordingly based on the taken image of the area to be processed. Thus, the control signal may be adapted to the situation captured by the image when processed at a specific location of the field.

As used herein, the term "identifying" includes detecting the status of an object, in other words, knowing that an object is at a certain location but not exactly what the object is, and identifying the status of the object, in other words, knowing the type of object that has been detected, in particular the species of the plant, like crops or weeds, insects and/or pathogens. In particular, the identification may be based on image identification and classification algorithms, such as convolutional neural networks or other algorithms known in the art. In particular, the identification of the object is location specific, which depends on the location of the processing device. In this way, the processing can be adapted in real time to the local situation in the field.

As used herein, the term "application rate decision logic" refers to a set of parameters provided to a processing device for controlling the processing device for treating a plant. In one embodiment, the application rate decision logic provides logic to decrease or increase the application rate depending on the expected loss of efficacy. The application rate decision logic preferably relates to a configuration file for the processing means. In other words, the application rate decision logic may be a decision tree with one or more layers for determining the control signals for controlling the processing means depending on measurable input variables, such as captured images and/or online field data. The application rate decision logic may include one layer relating to on/off decisions and optionally a second layer relating to the composition of the treatment product intended for use and further optionally a third layer relating to the application rate of the treatment product intended for use. In this manner, real-time decisions regarding the context of the processing are based on real-time images and/or online field data collected as the processing device passes through the field. Providing application rate decision logic prior to performing the processing reduces computational time while enabling reliable determination of control signals for processing. The application rate decision logic or profile may include location specific parameters provided to the processing device, which may be used to determine the control signals.

In one layer, the application rate decision logic may include thresholds related to parameters derived from captured images and/or object recognition. Such parameters may relate to parameters derived from images associated with the identified object(s) and which are decisive for the processing decision. Further parameters may be derived from the online field data. If the derived parameter is, for example, below a threshold, the decision is to shut down or not process, which is decisive for the processing decision. The decision is to switch on or process if the derived parameter is, for example, above a threshold. In this manner, the control signal is determined based on the application rate decision logic and the identified object. In the case of weeds, the parameters derived from the image and/or the weeds identified in the image may be based on parameters representing weed coverage. Similarly in the case of pathogens, the parameters derived from the images and/or the pathogens identified in the images may be based on parameters indicative of pathogen infection. Further similarly, in the case of insects, the parameters derived from the image and/or the insects identified in the image may be based on parameters representing the number of insects present in the image.

Preferably, the processing means is provided with application rate decision logic or a profile, based on which the processing means controls the processing arrangement. In a further embodiment, the determination of the profile includes a determination of an application rate level of the treatment product to be applied. The application rate decision logic may include layers regarding the application rate of the treatment product that relate to parameters derived from image and/or object recognition. Further parameters may be derived from the online field data. In other words, based on the profile, the processing device is controlled based on real-time parameters of the plant field, such as captured images and/or online field data, as to which application rate of the treatment product should be applied. In a preferred embodiment, the application rate decision logic comprises a variable or incremental application rate level dependent on one or more parameters derived from image and/or object recognition. In a further preferred embodiment, determining the application rate level based on the identified subject comprises determining a subject type, a subject growth stage and/or a subject density. Here, object density refers to the density of objects identified in a certain area. The object type, object growth phase and/or object density may be parameters derived from image and/or object recognition from which variable or incremental application rates are determined. By enhanced object identification based on geometric object contours, the application rate can be adjusted more robustly without losing efficacy, as the object type, object growth phase, object density and/or different environmental conditions can be included in the determination of the application rate.

The application rate decision logic may include another layer regarding the composition of the treatment product intended for use. In such cases, application rate decision logic may be determined depending on the expected significant yield or quality impact on the crop, ecological impact, and/or cost of treating the product composition. Thus, based on the application rate decision logic, the decision of whether to treat the field, which treatment product composition should be employed, at which application rate level, achieves the best possible result with respect to efficiency and/or efficacy. The application rate decision logic can include a barrel recipe for a process product barrel system of the processing device. In other words, the treatment product composition may represent a treatment product amount provided in one or more barrels of the treatment device prior to performing the treatment. The mixture from the one or more barrels forming the treatment product may be controlled on the fly depending on the determined composition of the treatment product. The treatment product preferably comprises at least one active ingredient, a diluent for diluting the active ingredient, for example water, a safener for increasing the resistance of the crop to the treatment product, an auxiliary for improving the mixing and treatment properties of the treatment product, micronutrients and/or liquid fertilizers, such as urea ammonium nitrate, UAN, for increasing the effectiveness of the treatment product.

As used herein, the term "application rate" describes the amount of treatment product per area, for example one liter of treatment product per hectare. In addition, the amount of active ingredient per area is described by the term "dose rate", e.g. 0.1 litres of active ingredient per hectare. Thus, determining the application rate preferably comprises determining the dose rate. In other words, by varying the composition of the treatment product, the dosage rate can be adjusted without changing the overall application rate.

The term "efficiency" refers to the balance of the amount of treatment product applied and the amount of treatment product effective to treat a plant in a field of plants. How effectively the treatment is performed depends on environmental factors such as weather and soil.

The term "efficacy" refers to the balance of positive and negative effects of the treatment product. In other words, efficacy refers to the optimum value of the application rate of the treatment product required to effectively treat a particular plant. The application rate should not be so high that the treatment product is wasted, which would also increase costs and negative effects on the environment, but not so low that the treatment product is not effectively treated, which may result in the plant being immunized against the treatment product. The efficacy of the treatment product also depends on environmental factors such as weather and soil.

As used herein, the term "treatment product" refers to products used in plant treatment, such as herbicides, insecticides, fungicides, plant growth regulators, nutritional products, and/or mixtures thereof. The treatment product may comprise different components including different active ingredients such as different herbicides, different fungicides, different insecticides, different nutritional products, different nutrients, and other components such as safeners (particularly in combination with herbicides), adjuvants, fertilizers, adjuvants, stabilizers, and/or mixtures thereof. The term "treatment product composition" is a composition comprising one or two or more treatment products. Thus, there are different types of e.g. herbicides, insecticides and/or fungicides, which are based on different active ingredients, respectively. Since the plant to be protected by the treatment product is preferably a crop, the treatment product may be referred to as a crop protection product. The treatment product composition may also comprise additional substances mixed with the treatment product, such as, for example, water, in particular for diluting and/or thinning the treatment product; and/or nutrient solutions, particularly for enhancing the efficacy of the treatment product. Preferably, the nutrient solution is a nitrogen-containing solution, such as liquid Urea Ammonium Nitrate (UAN).

As used herein, the term "nutritional product" refers to any product that is beneficial to plant nutrition and/or plant health, including but not limited to fertilizers, macronutrients, and micronutrients.

The application rate decision logic is preferably used to increase or decrease the application rate depending on the expected loss of efficacy of the treatment product for plant treatment.

Lowering the application rate, also referred to as application rate, of the treatment product increases efficacy, but also increases the risk of drug resistance if the application rate of the treatment product is insufficient. Thus, the application rate decision logic and/or decision factors are validated and adjusted accordingly.

Variable application rates can be used allowing for higher differentiation of application rates, taking into account subject species, subject growth stage, subject density and/or different environmental conditions. The application rate may be variably adjusted to achieve the target efficacy, for example, in increments ranging from 1% to 20%, more preferably in increments ranging from 2% to 15%, most preferably in increments ranging from 5% to 10%, more preferably in increments of 10%. However, every resolution of application rate is possible. Thus, the efficiency of treatment product application is significantly improved, thereby achieving the same target efficacy using less treatment product. More preferably, the application rate should not exceed an upper limit determined by the maximum dosage rate associated with the respective active ingredient of the treatment product, which is legally permissible according to applicable regulatory laws and regulations. Thus, preferably, the application rate decision logic does not increase the application rate if the application rate will exceed the upper limit.

Preferably, the processing means is provided with a profile based on which the processing means controls the processing arrangement. The application rate of the treatment arrangement is controlled based on the profile.

Preferably, the processing means are provided with the vat recipe based on regional knowledge and/or field specific information, in particular from previous reconnaissance and/or management information, preferably including soil information and/or farming information and/or sowing information and/or weather information.

Accordingly, an improved method for plant treatment of plants having variable application rates is provided.

In a preferred embodiment, capturing plant images of a field of plants, identifying objects on the captured images, determining application rates, and determining control signals for controlling the processing arrangement are performed as a real-time process such that the processing device is instantaneously controllable based on the captured images of the field of plants as the processing device progresses through the field at specific locations of the field.

Accordingly, an improved method for plant treatment of plants having variable application rates is provided.

In a preferred embodiment, the application rate decision logic provides logic for determining the application rate for treating a plant depending on the expected loss of efficacy of the crop to be cultivated in the field of plants.

Accordingly, an improved method for plant treatment of plants having variable application rates is provided.

In a preferred embodiment, the application rate decision logic comprises a variable application rate dependent on one or more parameters derived from image and/or object recognition.

Preferably, the off-line field data comprises local yield expectation data, resistance data relating to the likelihood of resistance of the plant to the treatment product, expected weather data, expected plant growth data, regional information data relating to different regions of the plant field, e.g. as determined based on biomass, expected soil data and/or legal restrictions data.

In a further embodiment, expected weather data refers to data reflecting forecasted weather conditions. Based on such data, the determination of a parameter package or profile of a treatment arrangement for administration is enhanced, as the efficacy impact on the treatment product may be included in the activation decision and the application rate. For example, if high humidity weather exists, a decision can be made to apply the treatment product, as it is very effective in such a case. The expected weather data may be spatially resolved to provide weather conditions in different areas or at different locations in the field of plants where processing decisions are to be made.

In further embodiments, the expected weather data includes various parameters such as temperature, UV intensity, humidity, rainfall forecast, evaporation, dew. Based on such data, the determination of a parameter package or profile of a treatment arrangement for administration is enhanced, as the efficacy impact on the treatment product may be included in the activation decision and the application rate. For example, if high temperature and high UV intensity are present, the application rate of the treated product can be increased to compensate for faster evaporation. On the other hand, if, for example, the temperature and UV intensity are moderate, the metabolism of the plants is more active and the application rate of the treated product can be reduced.

In further embodiments, it is contemplated that soil data, such as soil moisture data, may be accessed from an external repository. Based on such data, the determination of a parameter package or profile of a treatment arrangement for administration is enhanced, as the efficacy impact on the treatment product may be included in the activation decision and the application rate. For example, if high soil moisture is present, decisions may be taken not to apply treatment products due to sweeping effects. It is contemplated that the soil data may be spatially resolved to provide soil moisture characteristics in different regions or at different locations in a field of plants where processing decisions will be made.

In further embodiments, at least a portion of the offline field data includes historical yield maps, historical satellite images, and/or spatially unique crop growth models. In one example, a performance map may be generated based on historical satellite images including, for example, field images for different points in one season of a plurality of seasons. Such a performance map allows for identifying a change in fertility in a field, for example by mapping a number of seasonal more fertile or less fertile areas.

Preferably, at least a portion of the offline field data is determined from historical yield maps, digital Power zones (Zoner Power zones), and/or spatially unique crop growth models.

Preferably, the expected plant growth data is determined depending on the amount of water still available in the soil of the plant and/or the expected weather data.

Accordingly, an improved method for plant treatment of plants having variable application rates is provided.

In a preferred embodiment, the method comprises:

objects are identified based on geometric object contours, and the geometry of plants (e.g., weeds) is modeled based on their species and/or their growth stage.

As used herein, the term "geometric object outline" refers to an outline that describes a geometric feature of an object, like the shape of a plant, insect, and/or pathogen at a particular growth stage of the plant, insect, and/or pathogen.

If the object and/or its growth phase cannot be identified based on the captured image alone, the geometric object contour is preferably used to approximate the different object types and object growth phases associated with the identified object.

Application rates are then determined based on the determined application rate decision logic and the object identified by using the approximation of the geometric object contour.

Thus, online or real-time recognition of objects and/or their growth stages on the captured image may be improved.

In a preferred embodiment, the method comprises:

determining the application rate based on the identified object includes determining an object type, an object growth stage, and/or an object density.

Preferably, identifying the object comprises identifying a plant, preferably a plant type and/or a plant size; insects, preferably insect type and/or insect size; and/or pathogen, preferably pathogen type and/or pathogen size.

Accordingly, an improved method for plant treatment of plants having variable application rates is provided.

In a preferred embodiment, the method comprises:

determining, by a processing device, online field data related to current conditions on a field of plants; and

determining a control signal depending on the determined application rate decision logic and the determined recognition objects and/or the determined on-line field data.

Accordingly, an improved method for plant treatment of plants having variable application rates is provided.

Determining online field data by the processing device may include sensors mounted on or placed in the field and received by the processing device.

In a preferred embodiment, the method comprises:

the online field data relates to current weather data, current plant growth data, and/or current soil data, such as soil moisture data.

In one embodiment, the current weather data is recorded in flight or on site. Such current weather data may be generated by different types of weather sensors mounted on the processing device or one or more weather stations placed in or near the field. Thus, current weather data may be measured during movement of the processing device over the field of plants. Current weather data refers to data reflecting weather conditions in a field of plants at a location where a processing decision is to be made. The weather sensor is for example a rain, UV or wind sensor.

In a further embodiment, the current weather data includes various parameters such as temperature, UV intensity, humidity, rainfall forecast, evaporation, dew. Based on such data, the determination of the configuration of the treatment device for administration is enhanced, as the efficacy impact on the treatment product may be included in the activation decision and the application rate. For example, if high temperature and high UV intensity are present, the application rate of the treated product can be increased to compensate for faster evaporation.

In a further embodiment, the online field data includes current soil data. Such data may be provided by soil sensors placed in the field, or it may be accessed from, for example, a repository. In the latter case, the current soil data may be downloaded onto a storage medium of the agricultural machine that includes the treatment gun(s). Based on such data, the determination of the configuration of the treatment arrangement for administration is enhanced, as the efficacy impact on the treatment product may be included in the activation decision and the application rate. For example, if high soil moisture is present, a decision may be taken not to apply treatment product due to sweeping effects.

In further embodiments, current or expected weather data and/or current or expected soil moisture data may be provided to a growth stage model to further determine the growth stage of the plant, weed, or crop plant. Additionally or alternatively, weather data and soil data may be provided to the disease model. Based on such data, the determination of the configuration of the treatment device (particularly the part of the treatment arrangement like a single nozzle) for application is enhanced, as the efficacy impact on the treatment product, such as for example weeds and crops that will grow at different rates during this time as well as after application, may be included in the activation decision and application rate. Thus for example the size of the weeds at the time of application or the stage of infection of the pathogen (seen or derived from the time of infection in our model) can be included in the activation decision and application rate.

Accordingly, an improved method for plant treatment of plants having variable application rates is provided.

In a preferred embodiment, the method comprises the steps of:

determining and/or providing validation data depending on a performance review of a treatment of the plant; and

the application rate decision logic is adjusted depending on the verification data.

Preferably, the performance review includes manual control of the application rate decision logic and/or automated control of the application rate decision logic. For example, manual control involves farmers observing a field of plants and answering questionnaires. In a further example, the performance examination is performed by capturing an image of a portion of a field of plants that has been processed and analyzing the captured image. In other words, performance reviews assess the efficiency of a treatment and/or the efficacy of a treated product after a plant has been treated. For example, if an already treated weed is still present, despite its treatment, a performance audit will include information that the application rate decision logic for that treatment did not achieve the goal of killing the weed.

Accordingly, an improved method for plant treatment of plants having variable application rates is provided.

In a preferred embodiment, the method comprises:

the application rate decision logic is adjusted using a machine learning algorithm.

The machine learning algorithms may include decision trees, naive bayes classification, nearest neighbors, neural networks, convolutional or recursive neural networks, generating countermeasure networks, support vector machines, linear regression, logistic regression, random forests, and/or gradient boosting algorithms. Preferably, the results of the machine learning algorithm are used to adjust the application rate decision logic.

Preferably, the machine learning algorithm is organized to process inputs with high dimensionality into outputs with much lower dimensionality. Such a machine learning algorithm is referred to as "intelligent" because it can be "trained". The algorithm may be trained using a record of training data. The record of training data includes training input data and corresponding training output data. The training output data of a training data record is the result that is expected to be produced by a machine learning algorithm given the training input data of the same training data record as input. The deviation between this expected result and the actual result produced by the algorithm is observed and evaluated by means of a "loss function". The loss function is used as feedback for adjusting parameters of the internal processing chain of the machine learning algorithm. For example, the parameters may be adjusted with an optimization objective that minimizes the value of the loss function of the result when all training input data is fed into the machine learning algorithm and the result is compared to the corresponding training output data. The result of this training is that a relatively small number of training data records are given as "ground truth," enabling machine learning algorithms to perform their work well for a large number of input data records many orders of magnitude higher.

Accordingly, an improved method for plant treatment of plants having variable application rates is provided.

According to a further aspect, a field manager system for a processing device for plant processing of a field of plants, comprising: an offline field data interface adapted to receive offline field data relating to expected conditions on a field of plants; a machine learning unit adapted to determine an application rate decision logic of the processing device in dependence of offline field data; and a decision logic interface adapted to provide application rate decision logic to a processing device as described herein.

In a preferred embodiment, the field manager system comprises a validation data interface adapted to receive validation data, wherein the machine learning unit is adapted to adjust the application rate decision logic in dependence of the validation data.

According to a further aspect, a treatment device for plant treatment of a field of plants, comprising: an image capturing device adapted to take an image of a plant; a decision logic interface adapted to receive application rate decision logic from a field manager system as described herein; a processing arrangement adapted to process the plant species in dependence on the received application rate decision logic; an image recognition unit adapted to recognize an object on the photographed image; a process control unit adapted to determine a control signal for controlling the process arrangement in dependence on the received application rate decision logic and the identified object; wherein the decision logic interface of the processing device is connectable to a decision logic interface of a field manager system as described herein, wherein the processing device is adapted to activate the processing arrangement based on a control signal of the processing control unit.

Thus, higher application rate differences are allowed in view of weed species, weed growth stage, weed density and/or different environmental conditions. The application rate may be variably adjusted to achieve the target efficacy, for example in increments ranging from 1% to 20%, more preferably in increments ranging from 2% to 15%, most preferably in increments ranging from 5% to 10%, even more preferably in increments of 10%. However, every resolution of application rate is possible. Thus, the efficiency of herbicide application is significantly improved, thereby using less herbicide to achieve the same target efficacy.

In a preferred embodiment, the processing means comprises: an online field data interface adapted to receive online field data relating to current conditions on the field of plants, wherein the process control unit is adapted to determine a control signal for controlling the process arrangement in dependence on the received application rate decision logic and the identified object and/or online field data.

Accordingly, an improved method for plant treatment of plants having variable application rates is provided.

In a preferred embodiment, the treatment device is designed as a smart nebulizer, wherein the treatment arrangement is a nozzle arrangement.

The nozzle arrangement preferably comprises several independent nozzles which can be controlled independently. Thus, higher application rate differences are allowed in view of weed species, weed growth stage, weed density and/or different environmental conditions at each nozzle. The application rate at each nozzle can be variably adjusted to achieve the target efficacy, for example in increments in the range of 1% to 20%, more preferably in increments in the range of 2% to 15%, most preferably in increments in the range of 5% to 10%, even more preferably in increments of 10%. However, every resolution of application rate is possible. Thus, the efficiency of herbicide application is significantly improved, thereby using less herbicide to achieve the same target efficacy.

According to a further aspect, the processing system comprises a field manager system as described herein and a processing device as described herein.

Advantageously, the benefits provided by any of the above aspects apply equally to all other aspects and vice versa. The above aspects and examples will be apparent from and elucidated with reference to the embodiments described hereinafter.

Drawings

Exemplary embodiments will be described in the following with reference to the following drawings:

FIG. 1 shows a schematic diagram of a plant treatment arrangement;

FIG. 2 shows a flow chart of a plant treatment method;

FIG. 3 shows a schematic view of a processing device on a field of plants; and

fig. 4 shows a schematic view of an image with a detected object.

Detailed Description

Fig. 1 shows a plant processing system 400 for processing plants of a plant field 300 by at least one processing device 200 controlled by a field manager system 100.

The processing device 200, preferably a smart sprayer, comprises a processing control unit 210, an image capturing device 220, an image recognition unit 230 and a processing arrangement 270 as well as an application rate decision logic interface 240 and an online field data interface 250.

The image capture device 220 includes at least one camera configured to capture an image 20 of a field of plants 300. The captured image 20 is provided to the image recognition unit 230 of the processing device 200.

The field manager system 100 includes a machine learning unit 110. Further, the field manager system 100 includes an offline field data interface 150, an application rate decision logic interface 140, and a validation data interface 160. The field manager system 100 may refer to a data processing element, such as a microprocessor, microcontroller, Field Programmable Gate Array (FPGA), Central Processing Unit (CPU), Digital Signal Processor (DSP), capable of receiving field data, for example, via a Universal Service Bus (USB), physical cable, bluetooth, or another form of data connection. A field manager system 100 may be provided for each processing device 200. Alternatively, the field manager system may be a central field manager system, such as a Personal Computer (PC), for controlling the plurality of processing devices 200 in the field 300.

The field manager system 100 is provided with offline field data Doff related to expected condition data for the plant field 300. Preferably, the off-line field data Doff includes local yield expectation data, resistance data relating to the likelihood of plant resistance to treatment products, expected weather condition data, expected plant growth data, regional information data relating to different regions of a field of plants, expected soil moisture data, and/or legal restrictions data.

The offline field data Doff is provided by an external repository. For example, the expected weather condition data may be provided by a weather station that provides weather forecasts. The weather station may also be a local weather station located on a plant field or processing device. Alternatively, the expected weather condition data may be provided by a service provider that preferably uses satellite data for predicting weather. In addition, the expected plant growth data is provided, for example, from a database storing different plant growth stages or from a plant growth stage model that reviews expected growth stages of crops, weeds, and/or pathogens depending on past field condition data. It is contemplated that plant growth data may alternatively be provided by a plant model that is substantially a digital twin of the respective plant, and the growth stage of the plant estimated, particularly depending on prior field data. Further, expected soil moisture data is determined, for example, depending on past, present, and expected weather condition data. The offline field data Doff may also be provided by an external service provider.

Depending on the offline field data Doff, the machine learning unit 110 determines the application rate decision logic 10. Preferably, the machine learning unit 110 knows the scheduled time for the treatment of the plant. For example, a farmer provides information to the field management system 100 that he plans to treat plants in a particular field the next day. The application rate decision logic 10 is preferably represented as a configuration file provided to the application rate decision logic interface 140 of the field manager system 100. Ideally, the application rate decision logic 10 is determined by the machine learning unit 110 on the same day that the processing device 200 is using the application rate decision logic 10. Via application rate decision logic interface 140, application rate decision logic 10 is provided to processing device 200, in particular to application rate decision logic interface 240 of processing device 200. For example, the application rate decision logic 10 in the form of a configuration file is updated into the memory of the processing device 200.

When the application rate decision logic 10 is received by the processing device 200, in particular the processing control unit 210, processing of the plants in the plant field 300 can be started.

The processing device 200 moves around the plant field 300 and detects and identifies objects 30, particularly crops, weeds, pathogens, and/or insects on the plant field 300.

Thus, the image capture device 200 continuously captures images 20 of the field of plants 300. The image 20 is provided to an image recognition unit 230, which image recognition unit 230 performs an image analysis on the image 20 and detects and/or recognizes the object 30 on the image 20. The object 30 to be detected is preferably a crop, weed, pathogen and/or insect. Identifying the object comprises identifying a plant, preferably a plant type and/or a plant size; insects, preferably insect type and/or insect size; and/or pathogen, preferably pathogen type and/or pathogen size. For example, it should be recognized that there is a difference between amaranthus retroflexus and crab, or between bees and locusts, for example. The object 30 is provided to the process control unit 210.

If the image recognition analysis detects the object 30 but is not able to recognize the object 30 and/or its growth stage, the image recognition unit 230 is provided with geometric object contours relating to the expected geometric appearance of the different plants. Due to many different factors like reflections, unexpected weather conditions and/or unexpected growth stages of plants, the image recognition unit 230 may not be able to recognize the object 30 and/or its growth stage.

The geometric object contour models the kind of object and/or the geometry of the growth phase. Based on the captured image 20 and the geometric object contour, the image recognition unit 230 is able to recognize the object 30 and/or its growth stage that could not be recognized without the geometric object contour.

The process control unit 210 is provided with application rate decision logic 10 in the form of a configuration file. The application rate decision logic 10 can be shown as a decision tree, where, based on input data, the treatment of the plant is decided on different layers of the decision and the application rate of the treatment product is decided. For example, in a first step it is checked whether the biomass of the detected weeds exceeds a predetermined threshold set by the application rate decision logic 10. The biomass of the weeds is generally related to the extent of weed coverage in the captured image 20. For example, if the biomass of the weeds is below 4%, it is decided not to treat the weeds at all. If the biomass of the weed is above 4%, a further decision is made. For example, in the second step, if the biomass of the weeds is higher than 4%, it is decided whether or not to treat the weeds depending on the soil moisture. If the soil moisture exceeds a predetermined threshold, it is still decided to treat weeds, and otherwise it is decided not to treat weeds. This is because herbicides used to treat weeds are more effective when the weeds are in a growth phase induced by high soil moisture. The application rate decision logic 10 has included information about the expected soil moisture. Since it has been raining for the past few days, it is predicted that the soil moisture is above a predetermined threshold and it will be decided to treat the weeds. However, the process control unit 210 is also provided by the online field data Don, in which case additional data is provided to the process control unit 210 from the soil moisture sensor. Thus, the decision tree for the profile will be decided based on the online field data Don. In an exemplary embodiment, the online field data Don includes information that the soil moisture is below a predetermined threshold. Therefore, it was decided not to treat weeds.

The process control unit 210 generates process control signals S based on the application rate decision logic 10, the identified objects and/or the online field data Don. Thus, the processing control signal S contains information whether the identified object 20 should be processed or not. The process control unit 210 then provides a process control signal S to the processing arrangement 270, which processing arrangement 270 processes the plant based on the control signal S. The treatment arrangement 270 comprises in particular a chemical point spray gun with different nozzles, which makes it possible to spray herbicides, plant growth regulators, insecticides and/or fungicides with high precision.

Thus, the application rate decision logic 10 is provided in dependence on the off-line field data Doff relating to the expected field conditions. Based on the application rate decision logic 10, the processing device 200 can decide which plant should be processed based solely on the identified objects in the field. Thus, the efficiency of the treatment and/or the efficacy of the treated product may be improved. To further improve the efficiency of the treatment and/or the efficacy of the treated product, the online field data Don may be used to include the current measurable conditions of the field of plants.

The provided processing arrangement 400 is additionally capable of learning. The machine learning unit 110 determines the application rate decision logic 10 depending on a given heuristic. After plant treatment based on the provided application rate decision logic 10, it is possible to verify the efficiency of the treatment and the efficacy of the treated product. For example, a farmer may provide field data for a portion of a field of plants that has been previously processed based on application rate decision logic 10 to field manager system 100. This information is called authentication data V. The validation data V is provided to the field manager system 100 via the validation data interface 160, providing the validation data V to the machine learning unit 110. The machine learning unit 110 then adjusts the application rate decision logic 10 or a heuristic for determining the application rate decision logic 10 from the verification data V. For example, verification data V indicates that weeds that have been treated based on application rate decision logic 10 are not killed, and the adjusted application rate decision logic 10 lowers the threshold to treat plants in one of the branches of the underlying decision tree.

Instead of the application rate decision logic 10 in the form of a configuration file provided by the external field manager system 100 to the processing device 200, the functionality of the field manager system 100 can also be embedded in the processing device 200. For example, a processing device with relatively high computing power can integrate the field manager system 100 within the processing device 200. Alternatively, all the described functions of the field manager system 100 and up to the determination of the control signal S by the processing device 200 may be calculated outside the processing device 200 (preferably in a cloud service). Thus, the processing device 200 is simply a "dumb" device for processing the plant species depending on the provided control signal S.

Fig. 2 shows a flow chart of an exemplary embodiment of a plant treatment method, in particular showing an implementation of determining an application rate.

The plant to be treated is a weed from a field of plants 300. The weeds will be treated with a treatment product like a herbicide. The crop plant planted in this example is soybean. The country of planting in this example is brazil. The weed spectrum known in the machine learning unit 110 for the application rate decision logic 20 is large crabgrass (saururus chinensis) and amaranthus retroflexus (porcine weeds).

Based on the rainfall events, the maximum growth phase of the weeds was modeled for both species and geometric object contours were derived to distinguish them. For example, the crab is expected to be much smaller than the redroot amaranth when treated. Based on these assumptions, application rates were calculated for treatment of crab grass with 1l/ha of herbicide and for treatment of redroot amaranth with 2l/ha of herbicide.

The user is provided with a profile for the treatment device 200 and a barrel formulation corresponding to the expected weeds.

On plant 300, the user verifies the presence of the crab and redroot amaranth with an image recognition application, such as xarvio sculling.

The user sets up the treatment apparatus 200 with a profile and fills the treatment apparatus 200 according to the tank recipe, in particular with a combination of water and herbicide. The processing apparatus 200 includes a processing arrangement 270 having a plurality of individual nozzles. The configuration file will tell the processing device 200 how to control each nozzle of the processing device 200 based on the process control signal S and in particular the GPS position and the online data from the sensors of the processing device 200.

In the treatment of weeds, the following steps are performed.

In step S10, application rate decision logic 10 is determined based on the offline field data Doff associated with expected conditions on the plant field 300. In step S20, an image 20 of a plant of the plant field 300 is captured. In step S30, the object 30 on the captured image 20 is identified.

In step S40, an application rate is determined based on the application rate decision logic 10 and the identified subject 30.

In this case, it is checked whether the biomass of the identified subject 30 is greater than 0.2% according to the application rate decision logic 10. The percentage value is related to the percentage of the captured image 20 covered by the identified object 30. If the biomass of the recognition object 30 is less than 0.2%, the herbicide application rate is set to 0l/ha because the recognition object 30 is too small to be effectively treated. If the biomass of the recognition object 30 is more than 0.2%, the kind of the recognition object 30 is checked. Identifying the biomass percentage of the subject 30 is directly related to the species and/or growth stage of the subject 30.

If the weed species, and thus the identified object 30, is large, the herbicide application rate is set to 1 l/ha. If the weed species is Amaranthus retroflexus, the herbicide application rate is set to 2 l/ha. If the weed species cannot be determined and is therefore undetectable, the geometric object contour is provided to the image recognition unit 230, in particular by the machine learning unit 110, and step S30 is re-performed, identifying the object 30 on the captured image 20.

Thereafter, the weed species are likely to be identified and the application rate is set according to the respective species. Alternatively, it was examined whether the species of weeds was most likely amaranthus retroflexus, and thus the application rate was set to 2l/ha of herbicide. If the weed species is unlikely to be Amaranthus retroflexus, the application rate is set to 1l/ha of herbicide.

According to the application rate decision logic 10, it is now checked whether the biomass of the identified weeds is greater than 4%. If the identified weed biomass is greater than 4%, the herbicide application rate is increased by 10%. Otherwise, the application rate of the herbicide is unchanged.

According to the application rate decision logic 10, it is now checked whether the environmental humidity of the field of plants 300 is greater than 50%. If the humidity is greater than 50%, the application rate of the herbicide is increased by 10%. Otherwise, the application rate of the herbicide is unchanged.

According to the application rate decision logic 10, it is now checked whether an intensive cultivation system is present on the field of plants 300. In this case, the application rate of the herbicide was reduced by 5%. Otherwise, the application rate of the herbicide is unchanged.

Furthermore, in step S50, a control signal S for controlling the processing arrangement 50 of the processing device 200 is determined on the basis of the determined application rate.

Fig. 3 shows a processing device 200 in the form of an Unmanned Aerial Vehicle (UAV) flying through a field of plants 300 containing a crop 410. There are also many weeds 420 between crops 410, weeds 420 are particularly toxic, produce many seeds and may significantly affect crop yield. The weed 420 should not be tolerated in the field of plants 300 containing the crop 410.

UAV 200 has an image capture device 220 that includes one or more cameras and captures images as it flies through the plant field 300. UAV 200 also has a GPS and inertial navigation system that enables the position of UAV 200 to be determined and the orientation of camera 220 to be determined as well. From this information, the footprint of the image on the ground may be determined so that a particular portion of the image, such as an example of a crop, weed, insect, and/or pathogen type, may be located relative to absolute geospatial coordinates. The image data captured by the image capturing device 220 is transferred to the image recognition unit 230.

The image captured by the image capturing device 220 has a resolution that enables one type of crop to be distinguished from another type of crop, and a resolution that enables one type of weed to be distinguished from another type of weed, and a resolution that enables not only detection of insects but also distinguishing of one type of insect from another type of insect, and a resolution that enables one type of pathogen to be distinguished from another type of pathogen.

The image recognition unit 230 may be external to the UAV 200, but the UAV 200 itself may have the necessary processing capabilities to detect and identify crops, weeds, insects, and/or pathogens. The image recognition unit 120 processes the images using a machine learning algorithm, e.g., an artificial neural network that has been trained on many image examples of different types of crops, weeds, insects, and/or pathogens, to determine which object is present and also to determine the type of object.

The UAV also has a treatment arrangement 270, in particular a chemical point spray gun with different nozzles, which enables it to spray herbicides, plant growth regulators, insecticides and/or fungicides with high precision.

As shown in fig. 4, the image capture device 220 captures an image 20 of a field 300. The image recognition analysis detects four objects 30 and identifies a crop 410 (triangle), a first undesirable weed 420, an amaranthus retroflexus (circle), and a second undesirable weed 430, a crabgrass (diamond). However, in addition to this, unidentified weeds 440 (crossing) were also detected. Therefore, the image recognition unit 230 uses the geometric object contour provided by the machine learning unit 110 in order to recognize the weeds.

Accordingly, the efficiency of the image capturing apparatus 220 may be improved.

Reference mark

10 application rate decision logic

20 images

30 objects on an image

100 field manager system

110 machine learning unit

140 decision logic interface

150 offline field data interface

160 authentication data interface

200 processing device (UAV)

210 process control unit

220 image capturing device

230 image recognition unit

240 decision logic interface

250 online field data interface

270 processing arrangement

300 field of plants

400 processing system

410 crops

420 weeds (Amaranthus retroflexus)

430 weeds (Tang)

440 unidentified weeds

S processing control signals

Don online field data

Doff offline field data

V authentication data

S10 determining application rate decision logic

S20 shooting image

S30 recognizing an object on an image

S40 determining application rate

S50 determining a control signal