阅读说明：本技术 一种基于人工智能的病虫害防治系统 (Disease and pest control system based on artificial intelligence ) 是由 汤哲 陈正云 杨家铃 齐芳 于 2020-12-29 设计创作，主要内容包括：本发明公开了一种基于人工智能的病虫害防治系统,包括无线视觉前端、通信模组和云端服务器,无线视觉前端包括太阳能光伏板、太阳能充电控制电路、锂电池、MCU微处理器、GPS定位模组和图像采集模组,通过图像采集模组定时拍摄茶园已放置的粘虫板的高清图像,并将该高清图像发送至MCU微处理器,MCU微处理器通过通信模组将该高清图像在发送至云端服务器,云端服务器接收所述高清图像数据,并基于人工智能分析茶园虫害的种类及数量,并提供病虫害情况的预测预报,在此过程中,通过MCU微处理器控制太阳能充电控制电路实现太阳能光伏板对锂电池的充放电。具有自动化程度高、实时性强,且精准防治的优点。(The invention discloses an artificial intelligence-based pest control system, which comprises a wireless vision front end, a communication module and a cloud server, wherein the wireless vision front end comprises a solar photovoltaic panel, a solar charging control circuit, a lithium battery, an MCU (microprogrammed control Unit), a GPS (global positioning system) positioning module and an image acquisition module, high-definition images of pest sticking plates placed in a tea garden are regularly shot through the image acquisition module and are sent to the MCU, the MCU sends the high-definition images to the cloud server through the communication module, the cloud server receives the high-definition image data, analyzes the types and the number of pests in the tea garden based on artificial intelligence and provides prediction and forecast of pest conditions, and in the process, the solar photovoltaic panel is controlled by the MCU to charge and discharge the lithium battery. The method has the advantages of high automation degree, strong real-time performance and accurate prevention and treatment.)

1. The utility model provides a pest control system based on artificial intelligence, its characterized in that, includes wireless vision front end, communication module and high in the clouds server, pest control system regularly shoots the high definition image of the mythimna separata board that tea garden has placed through wireless vision front end, utilizes the high definition image data upload to the high in the clouds server of shooting of communication module, and the high in the clouds server is received high definition image data to kind and quantity based on artificial intelligence analysis tea garden pest, and provide the prediction forecast of the pest and disease damage condition, wherein:

the wireless vision front end is based on solar energy power supply, including solar photovoltaic board, solar charging control circuit, lithium cell, MCU microprocessor, GPS location module and image acquisition module, the solar photovoltaic board passes through solar charging control circuit and is connected with the lithium cell, MCU microprocessor is connected with solar charging control circuit, GPS location module and image acquisition module respectively, MCU microprocessor realizes the charge and discharge of solar photovoltaic board to the lithium cell through controlling solar charging control circuit, just MCU microprocessor still is connected with the high in the clouds server through the communication module.

2. A pest control system based on artificial intelligence according to claim 1, wherein the cloud server receives the high definition image data, analyzes the type and number of pests in the tea garden based on artificial intelligence, and provides prediction and forecast of pest conditions as follows:

s1, the cloud server receives a high-definition image of the pest sticking plate placed in the tea garden, which is shot at regular time through the wireless vision front end;

s2, establishing a disease and pest diagnosis model based on a feature extraction network and a multi-scale feature fusion network which introduce an attention mechanism on the cloud server;

s3, inputting the high-definition image into a disease and pest diagnosis model;

and S4, analyzing the type and the number of the pests in the tea garden by the pest diagnosis model according to the input high-definition images, and providing prediction and forecast of pest conditions.

3. A pest control system based on artificial intelligence according to claim 2, wherein the feature extraction network comprises a plurality of convolution layers, a pooling layer and an attention module, high-definition images are input into the feature extraction network, feature maps with target features of different resolutions are finally output through convolution of the convolution layers and pooling of the pooling layer, and then the feature maps are input into the multi-scale feature fusion network to integrate feature information of feature maps of each layer.

4. A pest control system based on artificial intelligence according to claim 2, wherein the feature extraction network introducing the attention mechanism performs feature adjustment of channel dimensions on the feature map obtained by forward propagation by using the attention mechanism, learns a set of weight parameters related to the channel dimensions of the feature map by backward propagation, and multiplies the feature map by the feature map in corresponding dimensions to obtain a feature map with higher discriminability of the high-definition image to be processed.

5. A pest control system based on artificial intelligence according to claim 2 wherein the multi-scale feature fusion network employs a cross-stage feature fusion method for channel connection to further enhance feature information, embodied as: in the multi-scale feature fusion network, lower-level feature maps have more positions and contour features of pests, higher-level feature maps have more semantic information, and a cross-stage feature fusion method fuses adjacent and non-adjacent feature maps simultaneously to obtain feature maps with richer feature information, so as to provide guarantee for the subsequent pest detection.

6. A pest control system based on artificial intelligence according to any one of claims 1 to 5 wherein when the voltage of the solar photovoltaic panel is higher than that of the lithium battery, the MCU microprocessor controls the solar charging control circuit to charge the lithium battery with the solar photovoltaic panel; when the electric quantity of the lithium battery reaches saturation, the MCU microprocessor controls the solar charging control circuit to further disconnect the solar photovoltaic panel to charge the lithium battery pack.

7. A pest control system based on artificial intelligence according to any one of claims 1 to 5 wherein the communication module includes at least one of a 5G network, a 4G network, a 3G network and a 2G network.

Technical Field

The invention relates to the technical field of artificial intelligence, in particular to a pest control system based on artificial intelligence.

Background

Insect pests are an important problem in agricultural production, and the development of modern agriculture in China is severely restricted. The traditional counting method based on machine vision needs workers to go deep into the field to take pictures and then transmit the pictures to a computer terminal for identification and counting, and is large in workload and poor in instantaneity.

Therefore, how to develop an artificial intelligence-based pest control system with high automation degree, strong real-time performance and accurate control becomes a problem to be solved by technical personnel in the field.

Disclosure of Invention

In view of the above, the invention provides an artificial intelligence-based pest control system, which regularly shoots high-definition images of pest sticking plates placed in a tea garden through a wireless vision front end based on solar power supply, uploads the high-definition images to a cloud server by using a communication module, and the cloud server analyzes the type and the number of pests in the tea garden according to the high-definition image data and by using an artificial intelligence technology, so as to provide prediction and forecast of pest conditions.

On one hand, the invention provides a hose device at the tail end of a boom, which comprises a wireless vision front end, a communication module and a cloud server, wherein a pest control system regularly shoots high-definition images of pest sticking plates placed in a tea garden through the wireless vision front end, the communication module is used for uploading the shot high-definition image data to the cloud server, the cloud server receives the high-definition image data, analyzes the types and the number of pests in the tea garden based on artificial intelligence and provides prediction and forecast of pest situations, and the pest control system comprises:

the wireless vision front end is based on solar energy power supply, including solar photovoltaic board, solar charging control circuit, lithium cell, MCU microprocessor, GPS location module and image acquisition module, the solar photovoltaic board passes through solar charging control circuit and is connected with the lithium cell, MCU microprocessor is connected with solar charging control circuit, GPS location module and image acquisition module respectively, MCU microprocessor realizes the charge and discharge of solar photovoltaic board to the lithium cell through controlling solar charging control circuit, just MCU microprocessor still is connected with the high in the clouds server through the communication module.

Further, the cloud server receives the high-definition image data, analyzes the type and the number of pests in the tea garden based on artificial intelligence, and provides prediction and forecast concrete performances of pest and disease conditions as follows:

s1, the cloud server receives a high-definition image of the pest sticking plate placed in the tea garden, which is shot at regular time through the wireless vision front end;

s2, establishing a disease and pest diagnosis model based on a feature extraction network and a multi-scale feature fusion network which introduce an attention mechanism on the cloud server;

s3, inputting the high-definition image into a disease and pest diagnosis model;

and S4, analyzing the type and the number of the pests in the tea garden by the pest diagnosis model according to the input high-definition images, and providing prediction and forecast of pest conditions.

Further, the feature extraction network comprises a plurality of convolution layers, a pooling layer and an attention module, high-definition images are input into the feature extraction network, feature graphs with different resolutions and target features are finally output through convolution of the convolution layers and pooling operation of the pooling layer, and then the feature graphs are input into the multi-scale feature fusion network to integrate feature information of feature graphs of all layers.

Further, the feature extraction network introducing the attention mechanism performs feature adjustment of channel dimensions on the feature map obtained by forward propagation by using the attention mechanism, learns a group of weight parameters related to the channel dimensions of the feature map by backward propagation, and multiplies the feature map by corresponding dimensions to obtain the feature map with higher discriminability of the high-definition image to be processed.

Further, the multi-scale feature fusion network adopts a cross-stage feature fusion method to perform channel connection to further enhance feature information, which is specifically represented as: in the multi-scale feature fusion network, the lower-level feature maps have more pest position and contour features, and the higher-level feature maps have more semantic information. The cross-stage fusion method fuses adjacent characteristic graphs and non-adjacent characteristic graphs simultaneously to obtain the characteristic graphs with richer characteristic information, and provides guarantee for the subsequent pest detection.

Further, when the voltage of the solar photovoltaic panel is higher than that of the lithium battery, the MCU microprocessor controls the solar charging control circuit to charge the lithium battery through the solar photovoltaic panel; when the electric quantity of the lithium battery reaches saturation, the MCU microprocessor controls the solar charging control circuit to further disconnect the solar photovoltaic panel to charge the lithium battery pack.

Further, the communication module at least comprises one of a 5G network, a 4G network, a 3G network and a 2G network.

The pest control system provided by the invention regularly shoots high-definition images of pest sticking plates placed in a tea garden through the image acquisition module, and sends the high-definition images to the MCU microprocessor, the MCU microprocessor sends the high-definition images to the cloud server through the communication module, the cloud server receives the high-definition image data, analyzes the types and the number of pests in the tea garden based on artificial intelligence and provides prediction and forecast of pest conditions, in the process, the GPS positioning module is used for positioning the position where the pest occurs, the solar photovoltaic panel is used for generating electricity, the lithium battery is used for supplying power, and meanwhile, the MCU microprocessor controls the solar charging control circuit to realize charging and discharging of the solar photovoltaic panel on the lithium battery. Compared with the prior art, the automatic control system cancels manual on-site photographing, and has the advantages of high automation degree, strong real-time performance and accurate control.

In a further technical scheme, a disease and pest diagnosis model is established on the basis of a feature extraction network and a multi-scale feature fusion network which introduce an attention mechanism on a cloud server, the multi-scale feature fusion network is connected with channels by adopting a cross-stage feature fusion method to further enhance feature information, the disease and pest diagnosis model analyzes the type and the number of tea garden pests according to input high-definition images, prediction and forecast of pest conditions are provided, and the accuracy of the disease and pest diagnosis model for detecting small target pests is further improved.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate an embodiment of the invention and, together with the description, serve to explain the invention and not to limit the invention. In the drawings:

fig. 1 is a block diagram of a pest control system based on artificial intelligence according to an embodiment of the present invention;

FIG. 2 is a flow chart of the cloud server for providing pest and disease damage condition prediction and forecast according to the present invention;

FIG. 3 is a schematic diagram of a cross-phase feature fusion method in a multi-scale feature fusion network according to the present invention;

FIG. 4 is a comparison of experimental results of the attention mechanism and cross-stage feature fusion method of the present invention on different network architectures;

FIG. 5 is a graph showing the predicted results of the disease and pest diagnosis model of the present invention.

Detailed Description

It should be noted that the embodiments and features of the embodiments may be combined with each other without conflict. The present invention will be described in detail below with reference to the embodiments with reference to the attached drawings.

As shown in fig. 1, a pest control system based on artificial intelligence, including wireless vision front end, communication module and high in the clouds server, pest control system regularly shoots the high definition image of the mythimna separata board that the tea garden has placed through wireless vision front end, utilizes the high definition image data that the communication module will shoot to reach the high in the clouds server, and the high in the clouds server is received high definition image data to kind and quantity based on artificial intelligence analysis tea garden pest, and provide the prediction forecast of the pest and disease damage condition, wherein:

the wireless vision front end is based on solar energy power supply, including solar photovoltaic panel, solar charging control circuit, lithium cell, MCU microprocessor, GPS location module and image acquisition module, solar photovoltaic panel passes through solar charging control circuit and is connected with the lithium cell, and MCU microprocessor is connected with solar charging control circuit, GPS location module and image acquisition module respectively, and MCU microprocessor realizes the charge and discharge of solar photovoltaic panel to the lithium cell through controlling solar charging control circuit, just MCU microprocessor still is connected with the high in the clouds server through communication module. It should be noted that, the MCU microprocessor controls the solar charging control circuit to charge and discharge the solar photovoltaic panel to the lithium battery through the following conditions: when the voltage of the solar photovoltaic panel is higher than that of the lithium battery, the MCU microprocessor controls the solar charging control circuit to charge the lithium battery through the solar photovoltaic panel; when the electric quantity of lithium cell reached the saturation, MCU microprocessor charges for lithium cell group through controlling solar charging control circuit and then disconnection solar photovoltaic board, and it has guaranteed greatly that the disease and pest control system also can normally work in rainy day, when improving the sustainable operating time of lithium cell, has also practiced thrift the power generation energy of solar photovoltaic board.

Above-mentioned pest control system regularly shoots the high definition image of the mythimna separata board that the tea garden has placed through the image acquisition module to with this high definition image transmission to MCU microprocessor, MCU microprocessor sends this high definition image to high in the clouds server through the communication module, the high in the clouds server is received high definition image data to kind and quantity based on artificial intelligence analysis tea garden pest, and provide the prediction forecast of the pest situation, at this in-process, GPS location module is used for fixing a position the position that the pest took place, solar photovoltaic board is used for the electricity generation, the lithium cell is used for supplying power for MCU microprocessor, realize the charge-discharge of solar photovoltaic board to the lithium cell through MCU microprocessor control solar charging control circuit simultaneously.

As a preferred embodiment of the present invention, as shown in fig. 2, the cloud server in the present invention specifically realizes prediction and forecast of pest and disease damage conditions through the following processes:

s1, the cloud server receives a high-definition image of the pest sticking plate placed in the tea garden, which is shot at regular time through the wireless vision front end;

s2, establishing a disease and pest diagnosis model based on a feature extraction network and a multi-scale feature fusion network which introduce an attention mechanism on the cloud server; by combining the multi-scale feature fusion network with the attention mechanism, the neural network can be promoted to focus more on the target, and the attention of the neural network on other unrelated objects is suppressed.

Preferably, the feature extraction network in this step includes a plurality of convolution layers, a pooling layer and an attention module, the high-definition image is input to the feature extraction network, feature maps with target features of different resolutions are finally output through convolution of the plurality of convolution layers and pooling operation of the pooling layer, and then each feature map is input to the multi-scale feature fusion network to integrate feature information of each layer of feature map; meanwhile, the feature extraction network introducing the attention mechanism adjusts the feature of the channel dimension of the feature graph obtained by forward propagation by using the attention mechanism, learns a group of weight parameters related to the channel dimension of the feature graph by backward propagation, and multiplies the feature graph by the corresponding dimension to obtain the feature graph with higher discriminability of the high-definition image to be processed.

S3, inputting the high-definition images in the step S1 into a disease and pest diagnosis model;

and S4, analyzing the type and the number of the pests in the tea garden by the pest diagnosis model according to the input high-definition images, and providing prediction and forecast of pest conditions.

Therefore, the pest sticking plate data image is detected by using the disease and pest diagnosis model, so that the working efficiency is improved, and the cost is reduced.

In addition, in a further technical scheme, as the pest target is smaller, the obtained features are fewer than those of a common detection target, and based on the problem, the multi-scale feature fusion network adopts a cross-stage feature fusion method to perform channel connection so as to further enhance feature information, namely, the fusion capability of the network to the features is further enhanced at a feature fusion layer, so that the accuracy of the pest diagnosis model for detecting the small target pest is improved. In the multi-scale feature fusion network, lower-level feature maps have more position and contour features of pests, higher-level feature maps have more semantic information, and a cross-stage feature fusion method fuses adjacent and non-adjacent feature maps simultaneously to obtain feature maps with more abundant feature information, so that guarantee is provided for detecting pests next time. Fig. 3 is a schematic diagram of a cross-stage feature fusion method in a multi-scale feature fusion network. As shown in fig. 3, in the multi-scale feature fusion network, the lower level feature maps C2 and C3 have more pest position and contour features, the higher level feature maps C4 and C5 have more semantic information, and the cross-stage feature fusion method is described as a dotted line part in fig. 3, which is illustrated by a feature map P3: in addition to the original feature fusion, the feature map P3 is obtained by adding and fusing a feature map obtained by down-sampling an adjacent feature map P4 and an nonadjacent feature map P5 to an output feature map C3 of a feature extraction network. Similarly, a cross-stage feature fusion strategy is applied to the bottom-up feature maps N4 and N5 to enhance the feature expression capability of different resolution feature maps for small target pests.

In order to verify the effectiveness of the attention mechanism and the cross-phase feature fusion method in the disease and pest diagnosis model, the experimental results are shown in fig. 4 by performing comparison experiments on a plurality of network models (ResNet-50, ResNet-50-D, ResNet-101, CSPDarknet53 and Darknet53), wherein 1 in fig. 4 represents a model without introducing the attention mechanism, 2 to a model with introducing the attention mechanism, and 3 represents a model with simultaneously introducing the attention mechanism and the cross-phase feature fusion method. As can be seen from FIG. 4, the model with attention mechanism and cross-stage feature fusion method is introduced simultaneously, so that the accuracy and recall rate are improved greatly with little parameter increase and running speed loss, and the model plays an important role in real-time detection and counting of pest detection models. FIG. 5 is a graph showing the predicted results of the disease and pest diagnosis model of the present invention.

It should be noted that the communication module in the present invention at least includes one of a 5G network, a 4G network, a 3G network, and a 2G network.

In summary, the invention has the following advantages:

(1) by arranging the solar photovoltaic panel and the lithium battery and controlling the solar charging control circuit through the MCU microprocessor, the charging and discharging of the solar photovoltaic panel to the lithium battery are realized, the normal operation of the pest control system in rainy days is greatly ensured, the sustainable working time of the lithium battery is prolonged, and meanwhile, the power generation energy of the solar photovoltaic panel is saved;

(2) a disease and pest diagnosis model is established on the basis of a feature extraction network and a multi-scale feature fusion network which introduce an attention mechanism on a cloud server, the multi-scale feature fusion network is connected with channels by adopting a cross-stage feature fusion method to further enhance feature information, the disease and pest diagnosis model analyzes the type and the number of tea garden pests according to input high-definition images, prediction and forecast of pest conditions are provided, and the accuracy of the disease and pest diagnosis model in detecting small target pests is further improved.

The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.