CN120410767A - A shrimp health status assessment method based on multi-source information fusion - Google Patents

A shrimp health status assessment method based on multi-source information fusion

Info

Publication number
CN120410767A
CN120410767A CN202510727874.5A CN202510727874A CN120410767A CN 120410767 A CN120410767 A CN 120410767A CN 202510727874 A CN202510727874 A CN 202510727874A CN 120410767 A CN120410767 A CN 120410767A
Authority
CN
China
Prior art keywords
prawn
image
influence coefficient
prawns
target
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202510727874.5A
Other languages
Chinese (zh)
Other versions
CN120410767B (en
Inventor
徐勇
崔正国
陈剑磊
崔鸿武
张旭志
曲克明
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Yellow Sea Fisheries Research Institute Chinese Academy of Fishery Sciences
Original Assignee
Yellow Sea Fisheries Research Institute Chinese Academy of Fishery Sciences
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Yellow Sea Fisheries Research Institute Chinese Academy of Fishery Sciences filed Critical Yellow Sea Fisheries Research Institute Chinese Academy of Fishery Sciences
Priority to CN202510727874.5A priority Critical patent/CN120410767B/en
Priority claimed from CN202510727874.5A external-priority patent/CN120410767B/en
Publication of CN120410767A publication Critical patent/CN120410767A/en
Application granted granted Critical
Publication of CN120410767B publication Critical patent/CN120410767B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/02—Agriculture; Fishing; Forestry; Mining
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/254—Fusion techniques of classification results, e.g. of results related to same input data
    • G06F18/256—Fusion techniques of classification results, e.g. of results related to same input data of results relating to different input data, e.g. multimodal recognition
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
    • G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
    • G06V10/464—Salient features, e.g. scale invariant feature transforms [SIFT] using a plurality of salient features, e.g. bag-of-words [BoW] representations
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Artificial Intelligence (AREA)
  • Data Mining & Analysis (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Multimedia (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Mathematical Physics (AREA)
  • Business, Economics & Management (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • Animal Husbandry (AREA)
  • Human Resources & Organizations (AREA)
  • Human Computer Interaction (AREA)
  • Marine Sciences & Fisheries (AREA)
  • Mining & Mineral Resources (AREA)
  • Geometry (AREA)
  • Economics (AREA)
  • Agronomy & Crop Science (AREA)
  • Marketing (AREA)
  • Primary Health Care (AREA)
  • Strategic Management (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Image Processing (AREA)

Abstract

本发明公开了一种基于多源信息融合的对虾健康状态评估方法，本发明涉及数据处理技术领域。包括以下步骤：获取不同生长阶段对虾图像并进行预处理，使用人工标记生成训练图像，提取对虾轮廓长度、面积和壳体对比度等特征；基于训练图像，建立目标识别和特征提取模型，训练后识别待监测养殖环境中的对虾并提取其行为特征参数；计算行为评估影响系数、生长评估影响系数，并获取水源特征参数以计算病发影响系数；最后综合生成健康状态评估指数，与健康判断阈值对比，输出健康状态判断结果，从而实现对虾健康状态的有效评估，实时监测和预警帮助养殖者及时采取措施，预防疾病的发生，提高对虾养殖的存活率和经济效益。

The present invention discloses a method for assessing the health status of shrimp based on multi-source information fusion, which relates to the field of data processing technology. The method comprises the following steps: obtaining and preprocessing shrimp images at different growth stages, generating training images using artificial labels, and extracting shrimp features such as contour length, area, and shell contrast; establishing a target recognition and feature extraction model based on the training images, and after training, identifying shrimp in the monitored aquaculture environment and extracting their behavioral characteristic parameters; calculating the behavioral assessment influence coefficient and the growth assessment influence coefficient, and obtaining water source characteristic parameters to calculate the disease incidence influence coefficient; and finally, comprehensively generating a health status assessment index, comparing it with a health judgment threshold, and outputting a health status judgment result. This method effectively assesses the health status of shrimp, and provides real-time monitoring and early warning to help farmers take timely measures to prevent the occurrence of diseases, thereby improving the survival rate and economic benefits of shrimp aquaculture.

Description

Multi-source information fusion-based prawn health status assessment method
Technical Field
The invention relates to the technical field of data processing, in particular to a method for evaluating the health state of prawns based on multi-source information fusion.
Background
In the global aquaculture industry, prawns are receiving a great deal of attention as a major economic aquatic product. Along with the progress of the cultivation technology, the scale of the shrimp cultivation is continuously enlarged, but at the same time, the health management problem faced in the cultivation process is increasingly prominent. Especially, the evaluation of the growth state and health condition of the prawns and the influence of environmental factors becomes an important subject for improving the yield and ensuring the safety of aquatic products. The traditional prawn health evaluation method is mostly dependent on manual experience and simple growth indexes, lacks systematicness and scientificity, is easily influenced by subjective factors, and causes inaccurate and unstable evaluation results.
In recent years, rapid development of computer vision and deep learning technology has brought new opportunities for aquaculture health management. By effectively fusing the multisource information, more accurate evaluation of the health state of the prawns can be realized. The image processing technology can provide detailed appearance characteristic information for the prawns, and the environment monitoring data provide background basis for the evaluation of the health state. However, how to integrate these diversified information efficiently and analyze and judge by scientific methods remains a great challenge in the current technical field.
Especially in the health monitoring of prawns, it is often necessary to consider both biological characteristics and environmental factors. For example, the swimming behavior, the physical characteristics of the prawns, the water quality index in the external environment, etc. may have a profound effect on the health status. The traditional method often cannot comprehensively consider the factors, so that the one-sided performance of the health state evaluation of the prawns is caused. Meanwhile, aiming at the dynamic changes of various environments and biological indexes, how to accurately acquire and process data in real time is also a problem to be solved in the prior art.
In the prior art, publication No. CN119474949A discloses a method for evaluating the health condition of the shrimp group based on deep cluster analysis, which comprises the steps of monitoring health parameters, determining a shrimp group dispersion grid, determining a temporary grid, identifying non-shrimp group abnormal grids, calculating an actual health index, adjusting the step length of the grid and evaluating the health. According to the invention, by combining physiological and behavioral data such as real-time density, saturation, liveness and the like, the health condition of the shrimp group can be accurately monitored and estimated, the abnormal grids are removed step by step, more representative grids are screened, the estimated deviation caused by environmental fluctuation or measurement errors is avoided, the dynamic monitoring of the health condition of the shrimp group is realized, and the grid step length is automatically adjusted through the comparison of an ideal health index and an actual health index, so that the accuracy and reliability of health estimation are improved. But the health of shrimp groups is affected by a variety of biological factors including physiological status, disease, nutritional levels, etc. Although this approach evaluates by focusing physiological and behavioral data, other uncontrollable biological factors may be ignored in the evaluation, affecting the overall assessment of health. Thus, the accuracy and the effectiveness of the evaluation result are reduced.
The above information disclosed in the background section is only for enhancement of understanding of the background of the disclosure and therefore it may include information that does not form the prior art that is already known to a person of ordinary skill in the art.
Disclosure of Invention
The invention aims to provide a multi-source information fusion-based prawn health state assessment method for solving the problems in the background technology.
In order to achieve the above purpose, the present invention provides the following technical solutions:
A method for evaluating health state of prawns based on multi-source information fusion comprises the following specific steps:
Obtaining a plurality of prawn images in different growth stages, carrying out image preprocessing on the collected prawn images, marking each prawn target in the preprocessed prawn images through a marking frame in a manual marking mode based on the preprocessed prawn images, obtaining image characteristic parameters corresponding to marked prawns in the images, and marking the marked images as training images, wherein the image characteristic parameters comprise the contour length, the contour area and the contrast of the prawn shells;
Based on the obtained training image, a target recognition model and a feature extraction model are established, the training image is used as the input of the target recognition model, the corresponding labeling frame in the training image is used as a label, the target recognition model is trained, the prawn area image in the labeling frame recognized by the target recognition model is cut to serve as the input of the feature extraction model, and the feature extraction model is trained by taking the corresponding image feature parameter as the label;
collecting continuous time sequence images of a shrimp culture environment to be monitored in an acquisition time period, preprocessing the continuous time sequence images, inputting the preprocessed continuous time sequence images into a target recognition model for training, recognizing shrimp targets in the images, adding a labeling frame to the images, inputting region images in the output labeling frame into a feature extraction model for training, obtaining image feature parameters corresponding to the shrimp targets, and extracting behavior feature parameters of each shrimp target in the acquisition time period through DeepSORT algorithm based on the generated labeling frame and the image feature parameters corresponding to the shrimp targets;
calculating and generating a behavior evaluation influence coefficient and a growth evaluation influence coefficient based on the obtained behavior characteristic parameters and image characteristic parameters of the prawn target, simultaneously obtaining water source characteristic parameters of the environment to be monitored for prawn culture in an acquisition time period, and calculating and obtaining a disease influence coefficient based on the water source characteristic parameters, wherein the behavior characteristic parameters comprise swimming displacement and swimming times of the prawn, and the water source characteristic parameters comprise the total number of water body colonies, the number of pathogenic vibrios, the water body temperature and the water body PH value;
And comprehensively generating a health state evaluation index according to the obtained disease influence coefficient, the behavior evaluation influence coefficient and the growth evaluation influence coefficient, comparing the generated health state evaluation index with a set health judgment threshold value, and sending out corresponding health state judgment results according to different comparison results to finish the health state evaluation of the prawns.
Further, acquiring a plurality of prawn images in different growth stages, and performing image preprocessing on the acquired prawn images, wherein the preprocessing comprises image enhancement and denoising preprocessing, wherein each prawn image is subjected to denoising processing by adopting a wavelet transformation denoising method, and each prawn image is subjected to image enhancement preprocessing by adopting bilateral filtering;
Marking each prawn target in the preprocessed prawn image through a marking frame in a manual marking mode, wherein the specific method of marking is that each prawn target in the image is marked through the marking frame by an image marking tool, and the marking frame is the smallest external rectangle of each prawn target.
Further, based on the obtained training image, a target recognition model is established, wherein the target recognition model is established based on a YOLOv s network model, the trained target recognition model is used for recognizing all the prawn targets in the prawn image for marking, and the marked region image is extracted;
Based on a convolutional neural network, a feature extraction model is established, wherein the convolutional neural network consists of an input layer, a convolutional layer, a pooling layer, a full-connection layer and an output layer, an activation function in the convolutional layer is a ReLU function, and the specific expression of the ReLU function is as follows:
ReLU(sl(p,q))＝max(0,sl(p,q))
Wherein l represents the first corresponding convolution layer, s l(p,q) represents the q-th eigenvalue of the p-th training sample image in the first corresponding convolution layer, where l is the index of the convolution layer, p is the index of the training sample image, and q is the index of the eigenvalue in the training sample image;
For the full-connection layer, setting the number of neurons of the full-connection layer as 32, setting the learning rate of the initial neural network as 0.001 and setting the training round number as 200;
The feature extraction model after training is input into the marked area image and output into the corresponding image feature parameter predicted value.
Further, calculating and generating a behavior evaluation influence coefficient and a growth evaluation influence coefficient based on the obtained behavior characteristic parameters and image characteristic parameters of the prawn target, wherein a formula based on the specific calculation of the behavior evaluation influence coefficient is as follows:
Wherein CIF is a behavioral assessment influence coefficient, V mean is an average speed of each swimming of all identified prawn targets in a prawn culture environment to be monitored in an acquisition time period, f mean is an average swimming number of all identified prawn targets in the prawn culture environment to be monitored in the acquisition time period, V 0 represents a set swimming speed reference value, f 0 is a set swimming number reference value, wherein the average speed V mean of each swimming of the prawns in the acquisition time period is calculated by displacement of each swimming and swimming time according to the following formula:
Wherein V u is the average speed of each swimming of the U-th prawn in the acquisition time period, U is the index of the identified prawn target, U is [1, U ], and U is the total number of the identified prawn targets.
Wherein X u,f is the displacement of the object of the u-th prawn in the F-th swimming, delta t u,f is the time interval of the object of the u-th prawn in the F-th swimming, F is the index of the swimming times of the object of the u-th prawn in the acquisition time period, F is the total swimming times of the object of the u-th prawn in the acquisition time period, and F is [1, F ];
The specific formula on which the growth evaluation influence coefficient is calculated is as follows:
Wherein ECI is a growth evaluation influence coefficient, AC mean is an average value of contrast of all identified prawn target shells in the prawn culture environment to be monitored, L mean is an average value of all identified prawn target contour lengths in the prawn culture environment to be monitored, M mean is an average value of all identified prawn target contour areas in the prawn culture environment to be monitored, and L 0 and M 0 are reference values of set prawn target contour lengths and prawn target contour areas.
Further, simultaneously acquiring water source characteristic parameters of the shrimp culture environment to be monitored in the acquisition time period, and calculating to obtain a disease incidence influence coefficient based on the water source characteristic parameters, wherein a formula specifically based on the disease incidence influence coefficient is calculated is as follows:
Wherein SCI is the disease incidence factor, RE is the total number of bacterial colonies in the water, HZ is the number of pathogenic vibrios, PH S is the PH value of the water, T S is the temperature of the water, PH 0 is the PH reference value of the water, and T 0 is the temperature reference value of the water.
Further, according to the obtained disease influence coefficient, behavior evaluation influence coefficient and growth evaluation influence coefficient, comprehensively generating a health state evaluation index, wherein the specific formula on which the health state evaluation index is calculated is as follows:
Where ZH is a health state assessment index, ω 1、ω2 and ω 3 are weight coefficients of a disease incidence influence coefficient, a behavior assessment influence coefficient and a growth assessment influence coefficient, respectively, where ω 1>ω2>ω3 and ω 1、ω2 and ω 3 are both greater than 0.
Further, comparing the generated health state evaluation index with a set health judgment threshold value, and sending out a corresponding health state judgment result according to different comparison results, wherein specific judgment logic is as follows:
When ZH is more than or equal to 1.0 xyz', judging that the health state of the prawns in the prawn culture environment to be monitored is optimal, wherein the current environment is suitable for the growth and development of the prawns;
When the ZH is less than or equal to 0.4 x yz 'and less than or equal to 1.0 x yz', judging that the health state of the prawns in the prawn culture environment to be monitored is good, indicating that the pathological change risk exists in the prawns in the current environment, and adjusting by adopting corresponding measures;
When the temperature is more than or equal to 0 and less than ZHH and less than 0.4 xyz', judging that the health state of the prawns in the prawn culture environment to be monitored is poor, and indicating that the prawns in the current environment have lesions and should be rearranged;
wherein yz 'is a health judgment threshold value, and is obtained by correcting the number of shrimps in the shrimp culture environment to be monitored, wherein the formula on which the health judgment threshold value yz' is calculated is as follows:
Wherein yz 0 is the initial value of the health judgment threshold, DS is the number of the prawns in the prawn culture environment to be monitored, and DS 0 is the reference number.
Compared with the prior art, the invention has the beneficial effects that:
Firstly, the method obtains key image characteristic parameters such as contour length, contour area and shell contrast by accurately labeling the prawn images in different growth stages. The extraction of the characteristics not only provides direct information of the growth state of the prawns, but also lays a foundation for subsequent target identification and characteristic extraction. By constructing the target recognition model and the feature extraction model, the automatic recognition and analysis of the individual prawns can be realized, the working efficiency is improved, and the subjective error of manual labeling is reduced. Secondly, the method can monitor the behavior characteristics of the prawns, such as swimming displacement and swimming times, in real time by analyzing continuous time sequence images of the culture environment. The process not only improves the timeliness of the health monitoring of the prawns, but also enhances the response capability to the change of the culture environment. The health state of the prawns can be comprehensively estimated by combining the characteristic parameters of the water source (such as the total number of water colonies, the number of pathogenic vibrios and the water quality index). Finally, based on comparison of the generated health state evaluation index and the set health judgment threshold, the health state judgment result can be sent out rapidly, and an effective early warning mechanism is formed. The capability of real-time monitoring and early warning can help breeders to take measures in time, prevent diseases and improve the success rate and economic benefit of prawn breeding.
Drawings
FIG. 1 is a schematic flow chart of the overall method of the present invention;
FIG. 2 is a graph of average rate-behavior evaluation influence coefficient fitting of prawns;
FIG. 3 is a graph showing the fit of the average swimming frequency-behavior evaluation influence coefficient of the prawns;
FIG. 4 is a graph of growth assessment influence coefficient calculation statistics;
FIG. 5 is a graph of contrast mean-growth evaluation influence coefficient fit;
FIG. 6 is a graph of statistics of disease incidence coefficient calculations;
FIG. 7 is a bar graph of a health status judgment evaluation of prawns;
FIG. 8 is a schematic diagram of identifying target criteria for prawns.
Detailed Description
The present invention will be further described in detail with reference to specific embodiments in order to make the objects, technical solutions and advantages of the present invention more apparent.
It is to be noted that unless otherwise defined, technical or scientific terms used herein should be taken in a general sense as understood by one of ordinary skill in the art to which the present invention belongs. The terms "first," "second," and the like, as used herein, do not denote any order, quantity, or importance, but rather are used to distinguish one element from another. The word "comprising" or "comprises", and the like, means that elements or items preceding the word are included in the element or item listed after the word and equivalents thereof, but does not exclude other elements or items. The terms "connected" or "connected," and the like, are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "up", "down", "left", "right" and the like are used only to indicate a relative positional relationship, and when the absolute position of the object to be described is changed, the relative positional relationship may be changed accordingly.
Examples:
referring to fig. 1-8, the present invention provides a technical solution:
A method for evaluating health state of prawns based on multi-source information fusion comprises the following specific steps:
step 1, acquiring a plurality of shrimp images in different growth stages, carrying out image preprocessing on the acquired shrimp images, marking each shrimp target in the preprocessed shrimp images through a marking frame in a manual marking mode based on the preprocessed shrimp images, acquiring image characteristic parameters corresponding to marked shrimps in the images, and marking the marked images as training images, wherein the image characteristic parameters comprise shrimp contour length, shrimp contour area and shrimp shell contrast.
Obtaining a plurality of prawn images in different growth stages, and carrying out image preprocessing on the collected prawn images, wherein the preprocessing comprises image enhancement and denoising preprocessing, wherein the denoising method adopting wavelet transformation is adopted to carry out denoising processing on each prawn image, and bilateral filtering is adopted to carry out image enhancement preprocessing on each prawn image;
the method for carrying out noise reduction and enhancement processing on the acquired prawn image comprises the following steps of carrying out noise reduction processing on the image after distortion correction by adopting a wavelet transformation denoising method, wherein the wavelet transformation denoising method comprises the specific steps of decomposing the image after distortion correction through wavelet transformation to obtain wavelet coefficients of the image in different scales and directions;
the acquired prawn image is subjected to detail enhancement by adopting bilateral filtering, and the specific formula based on filtering transformation is as follows:
Wherein y is a coordinate vector in an image coordinate system, I y is a gray value at the coordinate vector y, B y is a gray value obtained by bilateral filtering transformation of the gray value I y, G d and G r are gaussian functions, and the formulas according to G d and G r are as follows:
Where x is a coordinate vector in the image coordinate system, I x is a gray value at the coordinate vector x, and σ d and σ r are standard deviations of G d and G r, respectively.
Marking each prawn target in the preprocessed prawn image through a marking frame in a manual marking mode, wherein the specific method of marking is that each prawn target in the image is marked through the marking frame by an image marking tool, and the marking frame is the smallest external rectangle of each prawn target. Labeling is specifically performed by using LabelImg labeling tools, and after all labeling is completed, the labeling results are ensured to be stored in a proper format (such as XML, JSON, CSV) for subsequent model training and analysis.
The contour length of the prawn refers to the total length of the boundary of the object in the image, and can be generally obtained by using an edge detection algorithm, and the edge is detected by using a Canny edge detection method or a Sobel operator method, and then a contour detection function is used. The edge detection and contour detection methods are also used to obtain the contour area of the prawn.
The prawn shell contrast can be obtained by converting the image from the RGB color space to the HSV or Lab color space for better analysis of the color, and a histogram can be calculated using the functions in OpenCV, which can then be used to calculate the color contrast. The contrast of the prawn shell specifically refers to the difference value between the gray value of each pixel point and the gray value of the adjacent pixel points in the identified prawn target image, and the average value of the difference values of all the pixel points is taken as the contrast of the prawn shell.
And 2, based on the obtained training image, establishing a target recognition model and a feature extraction model, taking the training image as the input of the target recognition model, taking a corresponding labeling frame in the training image as a label, training the target recognition model, cutting a shrimp area image in the labeling frame recognized by the target recognition model to serve as the input of the feature extraction model, and taking a corresponding image feature parameter as a label, and training the feature extraction model.
Based on the obtained training image, a target recognition model is established, wherein the target recognition model is established based on YOLOv s network model, the trained target recognition model is used for recognizing all the prawn targets in the prawn image for marking, and the marked region image is extracted.
The YOLOv s network has the main function of real-time target detection. Object detection is an important task in the field of computer vision, whose object is to detect objects of various categories from images or videos and to accurately mark their positions. YOLOv5s is a model of YOLO (You Only Look Once) series, which balances the real-time performance and the accuracy, and is suitable for scenes with higher requirements on speed and lightweight models.
Based on a convolutional neural network, a feature extraction model is established, wherein the convolutional neural network consists of an input layer, a convolutional layer, a pooling layer, a full-connection layer and an output layer, an activation function in the convolutional layer is a ReLU function, and the specific expression of the ReLU function is as follows:
ReLU(sl(p,q))＝max(0,sl(p,q))
Wherein l represents the first corresponding convolution layer, s l(p,q) represents the q-th eigenvalue of the p-th training sample image in the first corresponding convolution layer, where l is the index of the convolution layer, p is the index of the training sample image, and q is the index of the eigenvalue in the training sample image;
For the full-connection layer, setting the number of neurons of the full-connection layer as 32, setting the learning rate of the initial neural network as 0.001 and setting the training round number as 200;
The feature extraction model after training is input into the marked area image and output into the corresponding image feature parameter predicted value.
And 3, collecting continuous time sequence images of the shrimp culture environment to be monitored in a collecting time period, preprocessing the continuous time sequence images, inputting the preprocessed continuous time sequence images into a target recognition model which is trained, recognizing shrimp targets in the images, adding a labeling frame to the images, inputting the region images in the output labeling frame into a feature extraction model which is trained, obtaining image feature parameters corresponding to the shrimp targets, and extracting behavior feature parameters of each shrimp target in the collecting time period through DeepSORT algorithm based on the generated labeling frame and the image feature parameters corresponding to the shrimp targets.
Preprocessing continuous time sequence images, wherein the preprocessing method is consistent with the preprocessing method, and is not repeated herein, wherein specific logic for extracting the behavior characteristic parameters of each prawn target in the acquisition time period through DeepSORT algorithm is that target detection is completed through tracking and identifying a labeling frame of each prawn output by a YOLOv s network, and after target detection, characteristic representation is extracted for each detected prawn, deepSORT. The feature extraction is obtained by using a convolutional neural network, the extracted features are used for subsequent matching and tracking, and DeepSORT uses a Kalman filter to predict and update the target state. Each object is modeled as a state variable including its position, velocity, etc. The kalman filter can predict the next position according to the motion model of the target in each frame so as to provide the motion track of the target, and in each frame DeepSORT can perform data association according to the result of target detection and the target position information of the previous frame through the hungarian algorithm (the data association refers to matching the detected target with the target tracked before). The algorithm will determine which newly detected objects are identical to which previous objects by calculating the distance between the detected object features and the existing object features, and DeepSORT will use the extracted features to determine the similarity between the objects by calculating euclidean distance or cosine similarity between the feature vectors during the data correlation process. Once the target is successfully tracked DeepSORT will continuously update the state of each shrimp and collect its behavioral characteristic parameters over time.
And 4, calculating and generating a behavior evaluation influence coefficient and a growth evaluation influence coefficient based on the obtained behavior characteristic parameters and image characteristic parameters of the prawn target, simultaneously obtaining the water source characteristic parameters of the prawn culture environment to be monitored in the acquisition time period, and calculating and obtaining the disease influence coefficient based on the water source characteristic parameters, wherein the behavior characteristic parameters comprise the swimming displacement and the swimming times of the prawn, and the water source characteristic parameters comprise the total number of water colonies, the number of pathogenic vibrios, the water temperature and the water PH value.
The specific acquisition method of the characteristic parameters of the water source comprises the steps of selecting a plurality of sampling points in the culture water body, and collecting the water sample by using a sterile sampler. During sampling, care should be taken to avoid external contamination of the sample, and the collected water sample is diluted, usually in a 10-fold dilution series. The diluted water samples are then inoculated onto a medium suitable for bacterial growth, such as Nutrient Agar (NA) or other selective medium, and the number of colonies cultivated on each medium is counted using plate Counting (CFU) to calculate the total number of colonies in the water sample.
Also, a plurality of sampling points are selected in the culture water body, a sterile sampler is used for collecting water samples, a specific selective culture medium such as TCBS agar is used for the growth of vibrio, the colony characteristics on the culture medium are observed, and the variety and the number of the vibrio are confirmed through a biochemical experiment or a molecular biological method (such as PCR).
Water temperature is measured directly in a body of water using a water thermometer such as a digital thermometer or a water quality monitoring instrument. Ensure that the probe of the thermometer is immersed in the water for a sufficient time to obtain an accurate reading.
Measurements were made using a portable pH meter or a laboratory pH meter. And (3) placing a probe of the pH meter into the water sample, and recording the pH value after the instrument is stable.
And calculating and generating a behavior evaluation influence coefficient and a growth evaluation influence coefficient based on the obtained behavior characteristic parameters and image characteristic parameters of the prawn target, wherein a formula based on the specific calculation of the behavior evaluation influence coefficient is as follows:
Wherein CIF is a behavior evaluation influence coefficient, V mean is an average speed of each swimming of all identified prawn targets in a prawn culture environment to be monitored in an acquisition time period, f mean is an average swimming number of all identified prawn targets in the prawn culture environment to be monitored in the acquisition time period, V 0 is a set swimming speed reference value, and f 0 is a set swimming number reference value. The average speed V mean of each swimming of the prawns in the collection time period is calculated through the displacement of each swimming and the swimming time, and the specific formula is as follows:
Wherein V u is the average speed of each swimming of the U-th prawn in the acquisition time period, U is the index of the identified prawn target, U is [1, U ], and U is the total number of the identified prawn targets.
Wherein X u,f is the displacement of the object of the u-th prawn in the f-th swimming, delta t u,f is the time interval of the object of the u-th prawn in the f-th swimming, f is the index of the swimming times of the object of the u-th prawn in the acquisition time period, and f is [1, F ].
It should be noted that, the behavior evaluation influence coefficient CIF is used for evaluating the health status of the prawn through the behavior parameters of the prawn, wherein the larger the behavior evaluation influence coefficient CIF value is, the better the health status of the prawn is.
The average swimming number f mean of the prawn in the collecting time period represents the activity frequency of the prawn in a certain time. When prawns are subjected to external stimuli (e.g., deterioration of water quality, temperature changes, hypoxia, predator threat, etc.), they typically exhibit more frequent swimming. This is a direct reaction to the environmental pressure, meaning that the prawn is struggling to evade the danger or to cope with an uncomfortable environment. The reduced number of swimming often means that the prawn may be ill or infected with parasites. In this case, the activity and function of the prawn are reduced, and the prawn is in a state of unwilling to move. Therefore, the difference value between the average swimming number f mean of the prawn in the acquisition time period and the set swimming number reference value is inversely proportional to the behavior evaluation influence coefficient CIF, and the influence of the swimming number is smoothed by ln1++ f mean-f0|/f0, so that the excessive amplification effect caused by the overhigh swimming number is avoided. The characteristic of the logarithmic function enables the value to change greatly in a small range, and the change gradually decreases in a large value, so that the characteristic of decreasing the marginal effect of the swimming times on health is reflected.
The average speed V mean of each swimming of all identified prawn targets in the prawn culture environment to be monitored in the acquisition time period represents the average speed of the swimming of the prawns. When prawns are stimulated, such as water pollution, temperature change, low-oxygen environment, disease infection and the like, the prawns generally show rapid swimming behavior. The rapid swimming is often a representation of stress, which means that the prawns are in an escape or defending state, the rapid swimming consumes a large amount of energy, the long-time stress and the high energy consumption can lead the prawns to be tired, influence the growth and reproductive capacity of the prawns and even possibly cause death, and the slow swimming usually means that the prawns can have health problems, such as diseases of bacteria, viruses or parasites, and the like, which can lead the activity of the prawns to be reduced, and if the water quality is poor, the prawns can become unqualified due to the influence of hypoxia or toxins, and the swimming is slow. Therefore, the difference between the average speed V mean of each swimming of the prawn in the acquisition time period and the set swimming speed reference value is inversely proportional to the behavior evaluation influence coefficient CIF through an exponential functionReflecting the significant impact of the swimming rate on the evaluation coefficient. Wherein the behavior evaluation influence coefficient calculation section data is as shown in table 1.
TABLE 1 statistical table of influence coefficient calculation for behavior evaluation
Through analysis of the data, a certain correlation exists between the behavior evaluation influence coefficient of the sample, the average speed and the average swimming times. For example, as can be seen from the data, the average rate showed a tendency to fluctuate as the sample number increased, and the average rate for sample number 7 was 7.0, showing a significant difference compared to 2.5 for sample number 11. This indicates that the average rate of the samples has a significant effect on the behavior evaluation influence coefficient, especially at higher rates, which is generally higher for the samples.
When the relation between the average running times and the influence coefficient of the behavior evaluation is analyzed, the average running times of the sample and the influence coefficient do not show obvious linear relation. For example, the average running number of sample No. 3 was 15, the behavior evaluation influence coefficient was only 0.33, and the average running number of sample No. 10 was 14, and the influence coefficient was 2.84.
The average swimming times f mean of all identified prawn targets in the prawn culture environment to be monitored in the acquisition time period are calculated through the swimming times of each prawn in the acquisition time period, meanwhile, the set swimming speed reference value V 0 represents the specific setting according to the length of the acquisition time period and the growth and development stage of the prawn in the current prawn culture environment to be detected, and the setting method of the swimming times reference value f 0 is the same.
The specific formula on which the growth evaluation influence coefficient is calculated is as follows:
Wherein ECI is a growth evaluation influence coefficient, AC mean is an average value of contrast of all identified prawn target shells in the prawn culture environment to be monitored, L mean is an average value of all identified prawn target contour lengths in the prawn culture environment to be monitored, M mean is an average value of all identified prawn target contour areas in the prawn culture environment to be monitored, and L 0 and M 0 are reference values of set prawn target contour lengths and prawn target contour areas.
It should be noted that, the growth evaluation influence coefficient ECI evaluates the health of the prawn through the growth state of the prawn, wherein the larger the growth evaluation influence coefficient ECI value is, the more healthy the growth of the prawn is.
The outline length of the prawn is generally directly related to the growth state of the prawn, and the prawn with over-large body length may grow too fast under certain specific conditions, so that the growth is generally unbalanced, the physical fitness of the prawn may be weaker, the resistance of the prawn is reduced, and the individual smaller prawn may be affected by diseases, so that the prawn cannot grow normally. Bacterial, viral or parasitic infections often cause stunted growth and development of the shrimp, so that the average value L mean of the target contour length of all identified shrimps in the shrimp culture environment to be monitored is inversely proportional to the difference between the reference values of the target contour lengths of the shrimps and the growth evaluation influence coefficient ECI, and the contour area can be regarded as an approximate measure of the volume of the shrimps. The problem that the outline area is overlarge and possibly has deformity exists is that the smaller outline area often indicates that the growth and development of the prawns are blocked, so that the difference value between the average value M mean of the identified target outline area of all the prawns and the reference value of the target outline area of the prawns in the culture environment of the prawns to be monitored is inversely proportional to the growth evaluation influence coefficient ECI, and the difference value is calculated byThe part combines the information of two dimensions of length and area, and adopts a square sum mode to quantify the growth characteristics of the prawns. The sum of squares can be regarded as a comprehensive measure of growth state in two dimensions, in which way the growth of the prawns can be reflected more comprehensively.
The average value of the identified target contour length and contour area of all the prawns in the prawn culture environment to be monitored is obtained by averaging the predicted value of the target contour length and contour area of all the prawns output by the model. The set reference values of the target contour length and the target contour area of the prawns are set according to the growth and development stages of the prawns in the current prawn culture environment to be detected and combined with expert experience.
If the contrast of the shell of the prawn is high, abnormal pigmentation, diseases, stress or malnutrition and other problems can be suggested to exist on the surface of the prawn. These health problems can cause the appearance of the shrimp to change, thereby affecting the color and texture of its shell, so that the contrast AC for of the shrimp shell is inversely proportional to the growth evaluation influence coefficient ECI, and the higher the contrast, the smaller the value of the growth evaluation influence coefficient is, reflecting the significant negative effect of the contrast on the growth evaluation. Wherein the growth evaluation influence coefficient calculation section data is as shown in table 2.
TABLE 2 statistical table of growth evaluation influence coefficient calculation
Meanwhile, acquiring water source characteristic parameters of the shrimp culture environment to be monitored in the acquisition time period, and calculating to obtain a disease incidence influence coefficient based on the water source characteristic parameters, wherein a formula specifically based on the disease incidence influence coefficient is calculated is as follows:
Wherein SCI is the disease incidence factor, RE is the total number of bacterial colonies in the water, HZ is the number of pathogenic vibrios, PH S is the PH value of the water, T S is the temperature of the water, PH 0 is the PH reference value of the water, and T 0 is the temperature reference value of the water.
The larger the value of the disease incidence factor SCI, the worse the growth environment of the prawn, and the worse the health state of the prawn, the possible existence of the diseased prawn.
Wherein, the total colony count RE of the water body represents the concentration of bacteria in the water body, especially the number of pathogenic bacteria. The higher the total number of colonies, the worse the sanitary condition of the water body is, and the risk of occurrence of diseases may be increased. Therefore, the total number RE of the water body colonies is in direct proportion to the disease incidence influence coefficient SCI, and the influence of small colony changes on the disease incidence influence coefficient can be smoothed through a logarithmic function ln (1+RE), and meanwhile numerical distortion caused by the fact that the total number of the colonies is extremely large is avoided.
The two items of |PH S-PH0|2+|TS-T0|2 are used for measuring the deviation between the current PH value and the temperature of the water body and the reference value, the pH value and the temperature are important parameters of the water quality, the deviation from the normal range can cause pressure on the health of the prawns, the possibility of disease occurrence is increased, and the influence of larger deviation is emphasized in a square mode, so that the disease occurrence influence coefficient is obviously increased if the water body parameter deviates from the reference value more.
The number HZ of pathogenic vibrios directly reflects the concentration of pathogenic vibrios in water, and the pathogenic vibrios are common pathogens in prawn culture, so that the number HZ of pathogenic vibrios is in direct proportion to the disease incidence coefficient SCI, the more the number of pathogenic vibrios is, the larger the disease incidence coefficient SCI is through an exponential function e -HZ, and the direct and obvious threat to the occurrence of diseases is reflected. Wherein the disease incidence coefficient calculation section data is shown in table 3.
TABLE 3 statistical table of disease incidence coefficient calculation
And 5, comprehensively generating a health state evaluation index according to the obtained disease influence coefficient, the behavior evaluation influence coefficient and the growth evaluation influence coefficient, comparing the generated health state evaluation index with a set health judgment threshold value, and sending out corresponding health state judgment results according to different comparison results to finish the health state evaluation of the prawns.
Comprehensively generating a health state evaluation index according to the obtained disease influence coefficient, the behavior evaluation influence coefficient and the growth evaluation influence coefficient, wherein the specific formula on which the health state evaluation index is calculated is as follows:
Where ZH is a health state assessment index, ω 1、ω2 and ω 3 are weight coefficients of a disease incidence influence coefficient, a behavior assessment influence coefficient and a growth assessment influence coefficient, respectively, where ω 1>ω2>ω3 and ω 1、ω2 and ω 3 are both greater than 0.
The health state evaluation index ZH is used to indicate the health state of the prawn, and the larger the value is, the higher the health state of the prawn is.
Because the above-mentioned relative relation between each influence coefficient and the health status assessment index ZH is not described herein, the small change of CIF can have relatively small influence on the health status assessment index by introducing the form of the logarithmic function ln (1+cif), and the larger CIF change can significantly influence the calculation of the index. This approach makes the health assessment less sensitive to slight behavioral changes, thereby focusing more on overall behavioral trends. Using square root formThe effect on extremes can be reduced, emphasizing small to moderate growth variations. Using square root formAnd the same is true.
Since the occurrence of diseases is often the main cause of death and economic loss of prawns in the cultivation process. Therefore, the disease affecting coefficient SCI is given the highest weight, the importance of the behavior evaluation affecting coefficient CIF is that it can reflect the activity level and appetite of the prawn, but its influence on the health state is smaller than the disease risk, and the growth evaluation affecting coefficient ECI reflects the growth state of the prawn, whose importance is not neglected, but its influence is relatively smaller when an acute disease occurs. Omega 1>ω2>ω3 is thus provided and omega 1、ω2 and omega 3 are both greater than 0.
Comparing the generated health state evaluation index with a set health judgment threshold value, and sending out corresponding health state judgment results according to different comparison results, wherein specific judgment logic is as follows:
When ZH is more than or equal to 1.0 xyz', judging that the health state of the prawns in the prawn culture environment to be monitored is optimal, wherein the current environment is suitable for the growth and development of the prawns;
When the ZH is less than or equal to 0.4 x yz 'and less than or equal to 1.0 x yz', judging that the health state of the prawns in the prawn culture environment to be monitored is good, indicating that the pathological change risk exists in the prawns in the current environment, and adjusting by adopting corresponding measures;
When ZH is more than or equal to 0 and less than 0.4 xyz', judging that the health state of the prawns in the prawn culture environment to be monitored is poor, and indicating that the prawns in the current environment have lesions and should be rearranged. Wherein the health state evaluation index calculation and health state judgment result part data are shown in table 4.
TABLE 4 statistical table of health status decisions
By analysis of the data, most of the samples showed a positive correlation between the higher health assessment index and good health judgment in the assessment. For example, the health status evaluation index of sample No. 4 is 0.3193, the behavior evaluation influence coefficient is 3.86, and the health status is judged to be excellent. This suggests that the samples perform well in health status assessment, probably due to their higher growth assessment impact coefficient (0.4505) and relatively lower disease impact coefficient (15.4083), thereby promoting their overall health status.
Furthermore, when analyzing the relationship of the disease impact coefficient to the health status assessment index, we found that a higher disease impact coefficient often corresponds to a lower health status assessment index. For example, the disease incidence factor for sample number 3 is as high as 10.2342, whereas the health status assessment index is 0.0653 only, and the final health status is judged to be poor. This suggests that an increase in the disease impact coefficient may significantly reduce the health status of the sample.
Finally, combining the results of the health state expert evaluation and the health state judgment, the high consistency between the expert evaluation and the evaluation index can be seen. For example, the health status evaluation index of sample No.14 is 0.2197, the expert evaluation is excellent, and the health status judgment is also excellent. The consistency indicates that the technical scheme can effectively reflect the actual health condition of the sample.
Wherein yz 'is a health judgment threshold value, and is obtained by correcting the number of shrimps in the shrimp culture environment to be monitored, wherein the formula on which the health judgment threshold value yz' is calculated is as follows:
Wherein yz 0 is the initial value of the health judgment threshold, DS is the number of the prawns in the prawn culture environment to be monitored, and DS 0 is the reference number.
In the shrimp culture, as the number of shrimps increases, the biological load in the water body also increases, resulting in the change of water quality (such as oxygen concentration, ammonia nitrogen concentration and the like). The change of water quality may generate greater pressure on the health of the prawns, and the greater the density of the prawns, the greater the pressure of competing resources (such as food, habitat space, etc.), which is easy to cause stress reaction and disease transmission. Therefore, the number DS of the prawns in the prawn culture environment to be monitored is in direct proportion to the health judgment threshold yz'.
The above formulas are all formulas with dimensions removed and numerical values calculated, the formulas are formulas with a large amount of data collected for software simulation to obtain the latest real situation, and preset parameters in the formulas are set by those skilled in the art according to the actual situation.
The above embodiments may be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented in software, the above-described embodiments may be implemented in whole or in part in the form of a computer program product. Those of skill in the art will appreciate that the elements and algorithm steps described in connection with the embodiments disclosed herein can be implemented as electronic hardware, or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution.
The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed over a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
The foregoing is merely illustrative of the present application, and the present application is not limited thereto, and any person skilled in the art will readily recognize that variations or substitutions are within the scope of the present application.

Claims (7)

1. A method for evaluating the health state of prawns based on multi-source information fusion is characterized by comprising the following specific steps:
Obtaining a plurality of prawn images in different growth stages, carrying out image preprocessing on the collected prawn images, marking each prawn target in the preprocessed prawn images through a marking frame in a manual marking mode based on the preprocessed prawn images, obtaining image characteristic parameters corresponding to marked prawns in the images, and marking the marked images as training images, wherein the image characteristic parameters comprise the contour length, the contour area and the contrast of the prawn shells;
Based on the obtained training image, a target recognition model and a feature extraction model are established, the training image is used as the input of the target recognition model, the corresponding labeling frame in the training image is used as a label, the target recognition model is trained, the prawn area image in the labeling frame recognized by the target recognition model is cut to serve as the input of the feature extraction model, and the feature extraction model is trained by taking the corresponding image feature parameter as the label;
collecting continuous time sequence images of a shrimp culture environment to be monitored in an acquisition time period, preprocessing the continuous time sequence images, inputting the preprocessed continuous time sequence images into a target recognition model for training, recognizing shrimp targets in the images, adding a labeling frame to the images, inputting region images in the output labeling frame into a feature extraction model for training, obtaining image feature parameters corresponding to the shrimp targets, and extracting behavior feature parameters of each shrimp target in the acquisition time period through DeepSORT algorithm based on the generated labeling frame and the image feature parameters corresponding to the shrimp targets;
calculating and generating a behavior evaluation influence coefficient and a growth evaluation influence coefficient based on the obtained behavior characteristic parameters and image characteristic parameters of the prawn target, simultaneously obtaining water source characteristic parameters of the environment to be monitored for prawn culture in an acquisition time period, and calculating and obtaining a disease influence coefficient based on the water source characteristic parameters, wherein the behavior characteristic parameters comprise swimming displacement and swimming times of the prawn, and the water source characteristic parameters comprise the total number of water body colonies, the number of pathogenic vibrios, the water body temperature and the water body PH value;
And comprehensively generating a health state evaluation index according to the obtained disease influence coefficient, the behavior evaluation influence coefficient and the growth evaluation influence coefficient, comparing the generated health state evaluation index with a set health judgment threshold value, and sending out corresponding health state judgment results according to different comparison results to finish the health state evaluation of the prawns.
2. The method for evaluating the health state of the prawns based on the multi-source information fusion of claim 1, which is characterized by obtaining a plurality of prawn images in different growth stages, and carrying out image preprocessing on the collected prawn images, wherein the preprocessing comprises image enhancement and denoising preprocessing, wherein each prawn image is subjected to denoising processing by adopting a wavelet transform denoising method, and each prawn image is subjected to image enhancement preprocessing by adopting bilateral filtering;
Marking each prawn target in the preprocessed prawn image through a marking frame in a manual marking mode, wherein the specific method of marking is that each prawn target in the image is marked through the marking frame by an image marking tool, and the marking frame is the smallest external rectangle of each prawn target.
3. The method for evaluating the health state of the prawns based on the multi-source information fusion according to claim 2, wherein a target recognition model is established based on the obtained training image, wherein the target recognition model is established based on a YOLOv s network model, and the trained target recognition model is used for recognizing all the prawn targets in the prawn image for marking and extracting the marked region image;
Based on a convolutional neural network, a feature extraction model is established, wherein the convolutional neural network consists of an input layer, a convolutional layer, a pooling layer, a full-connection layer and an output layer, an activation function in the convolutional layer is a ReLU function, and the specific expression of the ReLU function is as follows:
ReLU(sl(p,q))＝max(0,sl(p,q))
Wherein l represents the first corresponding convolution layer, s l(p,q) represents the q-th eigenvalue of the p-th training sample image in the first corresponding convolution layer, where l is the index of the convolution layer, p is the index of the training sample image, and q is the index of the eigenvalue in the training sample image;
For the full-connection layer, setting the number of neurons of the full-connection layer as 32, setting the learning rate of the initial neural network as 0.001 and setting the training round number as 200;
The feature extraction model after training is input into the marked area image and output into the corresponding image feature parameter predicted value.
4. The method for evaluating the health state of the prawns based on the multi-source information fusion of claim 1, wherein the method is characterized in that a behavior evaluation influence coefficient and a growth evaluation influence coefficient are generated based on the obtained behavior characteristic parameters and image characteristic parameters of the prawn target, and a formula based on the specific calculation of the behavior evaluation influence coefficient is as follows:
Wherein CIF is a behavioral assessment influence coefficient, V mean is an average speed of each swimming of all identified prawn targets in a prawn culture environment to be monitored in an acquisition time period, f mean is an average swimming number of all identified prawn targets in the prawn culture environment to be monitored in the acquisition time period, V 0 represents a set swimming speed reference value, f 0 is a set swimming number reference value, wherein the average speed V mean of each swimming of the prawns in the acquisition time period is calculated by displacement of each swimming and swimming time according to the following formula:
wherein V u is the average speed of each swimming of the U-th prawn in the acquisition time period, wherein U is the index of the identified prawn target, U is [1, U ], and U is the total number of the identified prawn targets;
Wherein X u,f is the displacement of the object of the u-th prawn in the F-th swimming, delta t u,f is the time interval of the object of the u-th prawn in the F-th swimming, F is the index of the swimming times of the object of the u-th prawn in the acquisition time period, F is the total swimming times of the object of the u-th prawn in the acquisition time period, and F is [1, F ];
The specific formula on which the growth evaluation influence coefficient is calculated is as follows:
Wherein ECI is a growth evaluation influence coefficient, AC mean is an average value of contrast of all identified prawn target shells in the prawn culture environment to be monitored, L mean is an average value of all identified prawn target contour lengths in the prawn culture environment to be monitored, M mean is an average value of all identified prawn target contour areas in the prawn culture environment to be monitored, and L 0 and M 0 are reference values of set prawn target contour lengths and prawn target contour areas.
5. The method for evaluating the health state of the prawns based on the multi-source information fusion of claim 4, wherein the method is characterized in that characteristic parameters of water sources of the culture environment of the prawns to be monitored in the acquisition time period are obtained simultaneously, and the disease influence coefficient is calculated based on the characteristic parameters of the water sources, wherein a formula on which the disease influence coefficient is calculated specifically is based is as follows:
Wherein SCI is the disease incidence factor, RE is the total number of bacterial colonies in the water, HZ is the number of pathogenic vibrios, PH S is the PH value of the water, T S is the temperature of the water, PH 0 is the PH reference value of the water, and T 0 is the temperature reference value of the water.
6. The method for evaluating the health state of the prawns based on the multi-source information fusion of claim 5, wherein the health state evaluation index is comprehensively generated according to the obtained disease influence coefficient, the behavior evaluation influence coefficient and the growth evaluation influence coefficient, and the specific formula on which the health state evaluation index is calculated is as follows:
Where ZH is a health state assessment index, ω 1、ω2 and ω 3 are weight coefficients of a disease incidence influence coefficient, a behavior assessment influence coefficient and a growth assessment influence coefficient, respectively, where ω 1>ω2>ω3 and ω 1、ω2 and ω 3 are both greater than 0.
7. The method for evaluating the health status of the prawns based on the multi-source information fusion of claim 6, wherein the generated health status evaluation index is compared with a set health judgment threshold value, and corresponding health status judgment results are sent out according to different comparison results, and specific judgment logic is as follows:
When ZH is more than or equal to 1.0 xyz', judging that the health state of the prawns in the prawn culture environment to be monitored is optimal, wherein the current environment is suitable for the growth and development of the prawns;
When the ZH is less than or equal to 0.4 x yz 'and less than or equal to 1.0 x yz', judging that the health state of the prawns in the prawn culture environment to be monitored is good, indicating that the pathological change risk exists in the prawns in the current environment, and adjusting by adopting corresponding measures;
When ZH is more than or equal to 0 and less than 0.4 xyz', judging that the health state of the prawns in the prawn culture environment to be monitored is poor, and indicating that the prawns in the current environment have lesions and should be rearranged;
wherein yz 'is a health judgment threshold value, and is obtained by correcting the number of shrimps in the shrimp culture environment to be monitored, wherein the formula on which the health judgment threshold value yz' is calculated is as follows:
Wherein yz 0 is the initial value of the health judgment threshold, DS is the number of the prawns in the prawn culture environment to be monitored, and DS 0 is the reference number.
CN202510727874.5A 2025-06-03 Multi-source information fusion-based prawn health status assessment method Active CN120410767B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202510727874.5A CN120410767B (en) 2025-06-03 Multi-source information fusion-based prawn health status assessment method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202510727874.5A CN120410767B (en) 2025-06-03 Multi-source information fusion-based prawn health status assessment method

Publications (2)

Publication Number Publication Date
CN120410767A true CN120410767A (en) 2025-08-01
CN120410767B CN120410767B (en) 2026-05-05

Family

ID=

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN121232580A (en) * 2025-12-04 2025-12-30 中国水产科学研究院南海水产研究所 A method for dynamic regulation of shrimp broodstock rearing environment based on multi-source data fusion

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117350435A (en) * 2023-10-30 2024-01-05 中国水产科学研究院南海水产研究所 Method and system for managing industrial prawn culture equipment
CN119515884A (en) * 2025-01-23 2025-02-25 广州机智云物联网科技有限公司 An automatic shrimp health assessment method based on multidimensional data
US12233327B1 (en) * 2024-06-24 2025-02-25 Digimithril Inc. System and method for officiating interference in sports, powered by artificial intelligence
CN119809131A (en) * 2024-12-30 2025-04-11 广东省水文局湛江水文分局 A reservoir water ecological health diagnosis method and system

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117350435A (en) * 2023-10-30 2024-01-05 中国水产科学研究院南海水产研究所 Method and system for managing industrial prawn culture equipment
US12233327B1 (en) * 2024-06-24 2025-02-25 Digimithril Inc. System and method for officiating interference in sports, powered by artificial intelligence
CN119809131A (en) * 2024-12-30 2025-04-11 广东省水文局湛江水文分局 A reservoir water ecological health diagnosis method and system
CN119515884A (en) * 2025-01-23 2025-02-25 广州机智云物联网科技有限公司 An automatic shrimp health assessment method based on multidimensional data

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
王斌;徐建瑜;王春琳;: "基于计算机视觉的梭子蟹蜕壳检测及不同背景对蜕壳的影响", 渔业现代化, no. 02, 20 April 2016 (2016-04-20), pages 14 - 19 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN121232580A (en) * 2025-12-04 2025-12-30 中国水产科学研究院南海水产研究所 A method for dynamic regulation of shrimp broodstock rearing environment based on multi-source data fusion

Similar Documents

Publication Publication Date Title
Qiongyan et al. Detecting spikes of wheat plants using neural networks with Laws texture energy
CN107316289B (en) Method for dividing rice ears in field based on deep learning and superpixel division
CN115067243B (en) A fishery monitoring and analysis method, system and storage medium based on Internet of Things technology
Whalley et al. Applications of image processing in viticulture: A review
CN116912025A (en) Comprehensive management method and system of livestock breeding information based on cloud-edge collaboration
CN107480721A (en) A kind of ox only ill data analysing method and device
CN119378943A (en) A method of coordinated increase of fishery resources based on large water surface environment
CN116798113A (en) A method for predicting potential disease in livestock
CN119251592A (en) A plant maturity classification and identification method, system, device and storage medium
CN118015062A (en) A livestock body size measurement method based on depth camera and instance segmentation algorithm
Kannan et al. Hybrid rater to quantify and measure the severity of infection and spread of infection in muskmelon
Nancy et al. Cucumber leaf disease detection using glcm features with random forest algorithm
Kumar et al. Detection of mastitis disease in cow with machine learning classifiers
CN120410767B (en) Multi-source information fusion-based prawn health status assessment method
Ueda et al. A Smartphone-Based Method for Assessing Tomato Nutrient Status Through Trichome Density Measurement
CN120223718A (en) A method and system for intelligent growth monitoring and analysis of marine aquaculture
CN120410767A (en) A shrimp health status assessment method based on multi-source information fusion
CN116523866B (en) Wheat scab resistance identification method, system, electronic equipment and storage medium
CN117292305B (en) Method, system, electronic equipment and medium for determining the number of fetal movements of fish fertilized eggs
Nirmala et al. Hybrid Deep Learning Method for Diagnosis of Cucurbita Leaf Diseases.
Felicetti et al. Fish blood cell as biological dosimeter: In between measurements, radiomics, preprocessing, and artificial intelligence
CN119147085A (en) System for living body estimation muscovy duck body chi
CN119068556B (en) A method for detecting vitality status of fish products
Vieira et al. An automatic method for estimating insect defoliation with visual highlights of consumed leaf tissue regions
CN114299064B (en) Pig health early warning system and method based on image recognition

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant