阅读说明：本技术 基于水下成像与高光谱技术的石斑鱼精准投饵及调控方法 (Accurate grouper feeding and regulating method based on underwater imaging and hyperspectral technology ) 是由 聂鹏程 曾国权 蔡铖勇 张慧 马建忠 何勇 於俊琦 于 2020-04-13 设计创作，主要内容包括：本发明公开一种基于水下成像与高光谱技术的石斑鱼精准投饵及调控方法,包括：(1)获取鱼肉样本的光谱与高光谱图像,建立光谱数据与鱼肉内待测物质的相关性模型；(2)改变饲料种类及投喂方式,利用高光谱检测鱼肉内待测物质的含量并建立含量与饲料种类及投喂方式的关系；(3)选取最佳鱼肉品质所对应的饲料品种及投喂方式对养殖水产进行投喂。本发明利用高光谱技术实现鱼肉品质的快速、无损在线检测的要求,水下成像技术分析鱼群摄食行为并建立养殖对象饥饿程度评估系统,根据检测分析结果以及鱼肉饥饿程度来智能调控投饵机实现精准投饵,具有重要的科学研究价值和应用意义。(The invention discloses a grouper accurate bait casting and regulating method based on underwater imaging and hyperspectral technology, which comprises the following steps: (1) acquiring a spectrum and a hyperspectral image of a fish sample, and establishing a correlation model of spectral data and a substance to be detected in fish; (2) changing the type of feed and the feeding mode, detecting the content of the substance to be detected in the fish meat by utilizing hyperspectrum and establishing the relation between the content and the type of the feed and the feeding mode; (3) and selecting a feed variety corresponding to the best fish quality and a feeding mode to feed the cultured aquatic products. The method utilizes the hyperspectral technology to realize the requirement of fast and nondestructive online detection on the fish quality, the underwater imaging technology analyzes the feeding behavior of fish schools and establishes a system for evaluating the hunger degree of the cultured objects, and the bait casting machine is intelligently regulated and controlled according to the detection and analysis result and the hunger degree of the fish to realize accurate bait casting, so that the method has important scientific research value and application significance.)

1. An accurate grouper feeding and regulating method based on underwater imaging and hyperspectral technology is characterized by comprising the following steps:

(1) acquiring a spectrum and a hyperspectral image of a fish sample, and establishing a correlation model of spectral data and a substance to be detected in fish;

(2) changing the type of feed and the feeding mode, detecting the content of the substance to be detected in the fish meat by utilizing hyperspectrum and establishing the relation between the content and the type of the feed and the feeding mode;

(3) and selecting a feed variety corresponding to the best fish quality and a feeding mode to feed the cultured aquatic products.

2. The method for accurately feeding and controlling the grouper based on the underwater imaging and hyperspectral technology as claimed in claim 1, wherein the nutrients comprise moisture, protein, fat, inorganic salts, minerals and vitamins.

3. The method for accurately feeding and regulating the grouper based on the underwater imaging and hyperspectral technology as claimed in claim 2, wherein the spectrum and the hyperspectral image of the fish sample are obtained by adopting visible/short wave near infrared 400-1100nm and near infrared 900-1700 nm.

4. The method for accurately feeding and regulating the grouper based on the underwater imaging and hyperspectral technology as claimed in claim 3, wherein the average spectrum of the sample image is extracted and a correlation model is established with the corresponding chemical value of the content of the substance to be measured by adopting partial least squares regression and least squares support vector machine.

5. The method for accurately feeding and controlling the groupers based on the underwater imaging and hyperspectral technology as claimed in claim 4, wherein 16 and 15 characteristic wavelengths are extracted in the visible/short wave near infrared and near infrared spectrum intervals respectively by adopting a competitive-based adaptive re-weighting algorithm, and PLSR and LS-SVM models are established respectively; and selecting an optimal model for predicting the content of the substance.

6. The method for accurately feeding and regulating the groupers based on the underwater imaging and hyperspectral technology as claimed in claim 1, wherein the underwater imaging is used for collecting images of the fish school, a space-time characteristic quantitative analysis model of the feeding behavior of the fish school is constructed, and the recursive neural network and physiological indexes of the fish school are combined to evaluate the whole hunger degree of the fish school.

7. The method for accurately feeding and regulating the grouper based on the underwater imaging and hyperspectral technology as claimed in claim 6, wherein the feeding device is controlled to feed the cultured object by the optimal feed obtained through hyperspectrum and a feeding mode, and the cultured object is detected in real time and the feeding device is intelligently regulated in real time by combining the starvation degree evaluation result of the cultured object by the underwater imaging system.

Technical Field

The invention relates to the technical field of accurate feeding and intelligent regulation and control of groupers, in particular to an accurate feeding and regulation and control method of groupers based on underwater imaging and hyperspectral technology.

Background

Circulating water is a novel high-density culture mode with higher popularization degree at home and abroad at present, and the mode can effectively relieve the challenge brought by the traditional culture mode to a certain extent. However, in actual production, feeding welfare of the cultured objects in the culture mode is difficult to guarantee, and starvation or over-satiation of the cultured objects is easily caused, so that the fish quality is affected. Moreover, the feed cost accounts for more than 40% of the aquaculture cost, and the economic benefit of enterprise breeding is directly influenced. The manual feeding in the traditional feeding method is time-consuming and labor-consuming, the cost is too high, the feeding efficiency is low, and the stability is poor; the machine feeds regularly and quantitatively, solves the problems of high cost, low feeding efficiency, poor stability and the like of a manual feeding mode to a certain extent, but easily causes insufficient or excessive feed feeding due to large difference between feeding time and feeding amount preset by a worker according to experience and actual feeding requirements of the cultured objects, and then causes reduction of growth efficiency of the cultured objects, feed waste and culture water pollution.

In recent years, researchers apply machine vision technology to the field of aquatic products, a new idea is provided for efficiently and quickly checking problems existing in the aquaculture process, and good progress is achieved. With the continuous innovation and development of modern analysis technologies such as image processing and spectroscopy, the detection of fish meat quality is developing towards the direction of rapidness, economy, accuracy and no damage. The hyperspectral imaging technology integrates the traditional spectrum technology (reflecting chemical composition and the like) and the image technology (reflecting morphological characteristics), not only can simultaneously capture the spectrum information and the image information of the tested sample, but also has the characteristics of high resolution, no need of pretreatment, non-destructiveness, high detection speed, easy operation and the like, overcomes the defect that the traditional spectrum technology cannot provide the spatial information of the tested sample, and breaks through the limitation that the image technology cannot provide the spectrum information. Currently, the hyperspectral imaging technology has obtained a great deal of research on nondestructive testing of aquatic products, and a lot of achievements are generated.

In conclusion, aiming at the problems of fixed bait feeding range, uneven feed feeding, low feed utilization rate and the like of a feeder in aquaculture, the intelligent feeding technology for high-density aquaculture based on underwater imaging and hyperspectral technology is developed.

Disclosure of Invention

The invention provides an accurate grouper feeding and regulating method based on underwater imaging and hyperspectral technology.

In order to achieve the above purpose, the invention adopts the following specific technical scheme:

an accurate grouper feeding and regulating method based on underwater imaging and hyperspectral technology comprises the following steps:

(1) acquiring a spectrum and a hyperspectral image of a fish sample, and establishing a correlation model of spectral data and a substance to be detected in fish;

(2) changing the type of feed and the feeding mode, detecting the content of the substance to be detected in the fish meat by utilizing hyperspectrum and establishing the relation between the content and the type of the feed and the feeding mode;

(3) and selecting a feed variety corresponding to the best fish quality and a feeding mode to feed the cultured aquatic products.

Preferably, the nutrients include moisture, protein, fat, inorganic salts, minerals and vitamins.

Preferably, the visible/short wave near infrared 400-1100nm and the near infrared 900-1700nm are adopted to obtain the spectrum and the hyperspectral image of the fish sample.

Preferably, the average spectrum of the sample image is extracted, and a correlation model is established by using partial least squares regression and a least squares support vector machine together with the corresponding chemical value of the content of the substance to be detected.

Preferably, 16 characteristic wavelengths and 15 characteristic wavelengths are respectively extracted in a visible/short wave near infrared and near infrared spectrum interval by adopting a competitive-based adaptive re-weighting algorithm, and PLSR and LS-SVM models are respectively established; and selecting an optimal model for predicting the content of the substance.

Preferably, underwater imaging is used for acquiring fish shoal images, a space-time characteristic of fish shoal feeding behaviors is constructed to carry out quantitative analysis model, and meanwhile, a recurrent neural network and fish shoal physiological indexes are combined to evaluate the whole hunger degree of the fish shoal.

Preferably, the feeding device is controlled to feed the breeding objects by the optimal feed obtained through hyperspectrum and a feeding mode, the breeding objects are detected in real time by combining the starvation degree evaluation result of the breeding objects by the underwater imaging system, and real-time intelligent regulation and control of the feeding device are realized.

The fish quality model based on the hyperspectral technology comprises different bait components and different feeding modes, and detects and models the content of nutrient substances in fish through hyperspectral detection.

The system for evaluating the hunger of the culture object comprises a model for quantitatively analyzing the space-time characteristics of the feeding behavior of the fish school based on computer vision, image processing and deep convolutional neural network technology, and is also suitable for semantization of the whole hunger degree of the fish school by combining a recurrent neural network and physiological indexes (gastrointestinal fullness, cortisol, amylase and the like) of the fish school.

On the basis of research of a system for evaluating the hunger degree of a culture object, CFD simulation and mechanical analysis technologies are combined, simulation and analysis are performed on all mechanical parts (a storage bin, a feed transmission part, a feed feeding part and the like) of a feeding device and connection structures among all the mechanical parts, and the feeding device is intelligently regulated and controlled to feed fish schools based on feeding welfare of the culture object.

The method utilizes the hyperspectral technology to realize the requirement of fast and nondestructive online detection on the fish quality, the underwater imaging technology analyzes the feeding behavior of fish schools and establishes a system for evaluating the hunger degree of the cultured objects, and the bait casting machine is intelligently regulated and controlled according to the detection and analysis result and the hunger degree of the fish to realize accurate bait casting, so that the method has important scientific research value and application significance.

Drawings

FIG. 1 is a schematic diagram of the method for accurately feeding and controlling Epinephelus;

FIG. 2 is a diagram of a visible-near infrared hyperspectral system;

FIG. 3 is a spectral plot of a fish sample;

fig. 4 is an image of fish flesh with different fat contents.

Detailed Description

The present invention will be described in detail with reference to the following examples and drawings, but the present invention is not limited thereto.

In the embodiment, the overall structure of the method for accurately feeding and regulating the grouper based on underwater imaging and hyperspectral technology is shown in fig. 1, and the method comprises the following steps:

the structure of the visible-near infrared hyperspectral system is shown in fig. 2, and mainly comprises a spectrometer (Impactor V10E, Specim, Spectral Imaging Ltd, Finland), a linear array CCD camera (Hamamatsu, Japan), a lens, a halogen light source, a conveyor belt controlled by a stepping motor, a computer and the like.

The fish meat is rich in nutrients such as water, protein, fat, inorganic salt, minerals and vitamins. The content and distribution of chemical components in fish meat are different, which may affect the quality of fish meat, for example, the moisture content and distribution in the muscle may affect the taste of fish meat, and the amount of fat content may be related to the tenderness of fish meat, as shown in fig. 4. The hyperspectrum belongs to frequency doubling and frequency combining spectrums of molecular vibration, is consistent with the absorption areas of frequency doubling and frequency combining of hydrogen-containing groups such as O-H, N-H, C-H and the like, and can be used for qualitatively and quantitatively analyzing the chemical components of a detected sample by extracting spectrum data and analyzing the characteristic information of the hydrogen-containing groups in the sample, wherein biochemical groups corresponding to different spectrum wavelengths are shown in table 1.

TABLE 1 Biochemical groups corresponding to different spectral wavelengths

A visible/short wave near infrared (400-1100nm) and near infrared (900-1700nm) system is adopted to obtain a spectrum and a hyperspectral image of a fish sample, as shown in FIG. 3. An average spectrum of a sample image is extracted, and a correlation model is established by Partial Least Squares Regression (PLSR) and least squares support vector machine (LS-SVM) according to the content chemical value of a corresponding measured substance (water, fat and protein). In order to reduce the collinearity and the redundancy of a hyperspectral image, 16 characteristic wavelengths and 15 characteristic wavelengths are respectively extracted in a visible/short wave near-infrared and near-infrared spectrum interval based on a competitive adaptive re-weighting algorithm (CARS), and PLSR and LS-SVM models are respectively established. Finally, the optimal model is used to predict the substance content. After the feed variety and the feeding mode are changed, the hyperspectral detection is used for detecting the content of the substance to be detected, the relation between the content and the feed variety and the feeding mode is established, and the feed variety and the feeding method corresponding to the optimal fish meat quality are selected for feeding the cultured aquatic products.

The system for evaluating the hunger of the breeding object comprises a model for quantitatively analyzing the time-space characteristics of the ingestion behavior of the fish school based on computer vision, image processing and deep convolutional neural network technology, and an evaluation system suitable for semantization of the whole hunger degree of the fish school is established by combining a recurrent neural network and physiological indexes (gastrointestinal fullness, cortisol, amylase and the like) of the fish school.

On the basis of research of a system for evaluating the hunger degree of a culture object, CFD simulation and mechanical analysis technologies are combined to simulate and analyze all mechanical parts (a storage bin, a feed transmission part, a feed feeding part and the like) of a feeding device and a connection structure among all the mechanical parts, and an intelligent feeding device mechanical mechanism based on the feeding welfare of the culture object is designed. Aiming at feeds with different grain diameters, the feed transmission part of the feeding device is dynamically simulated and optimized by utilizing a CFD simulation technology.

The optimal feed obtained through hyperspectrum and the feeding mode are used for controlling the feeding device to feed the cultured object, and the fed object is detected in real time by combining the starvation degree evaluation result of the cultured object with the underwater imaging system, so that the real-time intelligent regulation and control of the feeding device are realized.

The above description is only exemplary of the preferred embodiments of the present invention, and is not intended to limit the present invention, and any modifications, equivalents, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.