Fish feeding control method and system, electronic equipment and storage medium

Abstract

The invention provides a fish feeding control method and system, electronic equipment and a storage medium, and the method comprises the steps: collecting a target fish school video in a target fish school feeding state; according to the target fish school video, determining a dynamic feature and a static feature of a fish school ingestion behavior; wherein the dynamic feature is the confusion degree of the overall movement direction of the target fish school; the static feature is a texture feature of the target fish school ingestion video image; according to the dynamic characteristics and the static characteristics, determining a fish shoal feeding behavior state; wherein the ingestion behavior state of the fish school comprises a satiation state and a starvation state; if it is determined that the target fish school is in the satiation state, feeding is stopped. The fish feeding requirement can be accurately determined, real-time on-demand feeding of fishes is scientifically achieved, the bait utilization rate is effectively increased, and the culture benefits are increased.

Description

Technical Field

[0001] The present invention relates to the technical field of digital image processing, and in particular to a fish feeding control method and system, electronic equipment and storage medium. Background Art

[0002] In recent years, with the increasing demand for aquaculture, refined and intelligent aquaculture methods have gradually become the main development trend of the future aquaculture industry. The factory-scale recirculating aquaculture system has gradually replaced the traditional extensive (pond aquaculture) or semi-extensive (cage aquaculture) aquaculture methods with its ultra-high water resource utilization and aquaculture efficiency.

[0003] Since the precise control of the amount of fish fed in factory-scale recirculating aquaculture is a key link in refined farming, if accurate feeding cannot be achieved, it will inevitably cause feed waste and deterioration of the water quality environment, affecting the healthy growth of fish.

[0004] At present, the main methods of fish feeding are artificial observation feeding or simple mechanical feeding. The feeding is carried out at a fixed time and quantity according to the size of the fish and the breeding experience. It is easy to ignore the changes in the fish's feeding needs and the changes in the environment. It is impossible to accurately control the feeding amount according to the actual feeding behavior changes and growth conditions of the fish, which leads to problems such as feed waste, residual bait in the water polluting the water quality, etc., which are detrimental to the breeding benefits.

[0005] Therefore, how to provide a fish feeding control method and system, electronic equipment and storage medium to accurately determine the fish feeding needs, scientifically realize real-time on-demand feeding of fish, effectively improve bait utilization, and improve breeding efficiency has become an urgent problem to be solved. Summary of the invention

[0006] In view of the defects in the prior art, the embodiments of the present invention provide a fish feeding control method and system, an electronic device and a storage medium.

[0007] The present invention provides a fish feeding control method, comprising:

[0008] When the target fish school is feeding, collect the video of the target fish school;

[0009] According to the target fish school video, the dynamic characteristics and static characteristics of the fish school's feeding behavior are determined; the dynamic characteristics are the degree of confusion of the overall movement direction of the target fish school; the static characteristics are the texture characteristics of the target fish school's feeding video image;

[0010] Determine the feeding behavior state of the school of fish according to the dynamic characteristics and the static characteristics; wherein the feeding behavior state of the school of fish includes: a satiation state and a hunger state;

[0011] If it is determined that the target fish is in a state of fullness, stop feeding.

[0012] According to the fish feeding control method provided by the present invention, the dynamic characteristics and static characteristics of the feeding behavior of the fish school are determined according to the target fish school video, which specifically includes:

[0013] According to the target fish school video, determine several target fish school feeding video key frames;

[0014] Preprocess the key frames of the target fish school feeding video to determine the target fish school feeding image;

[0015] According to the target fish school feeding images, the dynamic and static characteristics of the fish school feeding are determined.

[0016] According to the fish feeding control method provided by the present invention, the key frames of the target fish school feeding video are preprocessed to determine the target fish school feeding image, specifically including:

[0017] The target fish feeding images are determined by processing the target fish feeding video key frames through homomorphic filtering and histogram equalization.

[0018] According to the fish feeding control method provided by the present invention, the dynamic characteristics and static characteristics of the fish school feeding are determined according to the target fish school feeding image, which specifically includes:

[0019] According to the LK optical flow method, the motion vectors of several target fish feeding images are calculated; wherein the motion vectors include the lateral motion speed and the longitudinal motion speed of the pixel points in the image;

[0020] Based on the motion vectors, the speed and the rotation angle of each motion vector are calculated;

[0021] According to the speed and angle of each motion vector, the discrete degree of speed and angle is calculated to determine the dynamic characteristics of fish feeding;

[0022] The texture features of the target fish feeding image are extracted based on the gray-level co-occurrence matrix, and the static characteristics of the fish feeding are determined.

[0023] According to the fish feeding control method provided by the present invention, the texture features include: angular second-order moment, inverse disparity, contrast, entropy and correlation.

[0024] According to the fish feeding control method provided by the present invention, the feeding behavior state of the fish school is determined according to the dynamic characteristics and the static characteristics, which specifically includes:

[0025] The dynamic features and static features are fused and reduced by the dimensionality reduction algorithm to determine the feeding fusion features of the target fish school;

[0026] The target fish school feeding fusion features are input into the trained target fish school feeding state recognition model to determine the feeding behavior state of the fish school.

[0027] According to the fish feeding control method provided by the present invention, before the step of inputting the target fish school feeding fusion feature into the trained target fish school feeding state recognition model to determine the feeding behavior state of the fish school, the method further includes:

[0028] Based on the support vector machine classifier, a fish feeding state recognition model is determined;

[0029] The radial basis function is selected as the kernel function of the support vector machine classifier, the fish feeding state recognition model is trained, and the target fish feeding state recognition model is determined.

[0030] The present invention also provides a fish feeding control system, comprising: a video acquisition unit, a feature determination unit, a state recognition unit and a feeding control unit;

[0031] A video acquisition unit, used for acquiring a video of a target fish school when the target fish school is feeding;

[0032] A feature determination unit is used to determine the dynamic features and static features of the feeding behavior of the fish school according to the target fish school video; wherein the dynamic feature is the degree of confusion of the overall movement direction of the target fish school; and the static feature is the texture feature of the target fish school feeding video image;

[0033] A state recognition unit is used to determine the feeding behavior state of the school of fish according to the dynamic characteristics and the static characteristics; wherein the feeding behavior state of the school of fish includes: a satiation state and a hunger state;

[0034] The feeding control unit is used to stop feeding if it is determined that the target fish school is in a state of fullness.

[0035] The present invention also provides an electronic device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of any of the above-mentioned fish feeding control methods are implemented.

[0036] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of any of the above-mentioned fish feeding control methods are implemented.

[0037] The fish feeding control method and system, electronic device and storage medium provided by the present invention determine the dynamic characteristics and static characteristics of the feeding behavior of the fish school through video image analysis when the target fish school is in a feeding state, and use the dynamic characteristics and static characteristics to scientifically detect and analyze the feeding behavior of the fish school, accurately determine the feeding needs of the fish, and use the analysis results to scientifically guide feeding, effectively improve the utilization rate of bait, and improve the breeding efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0039] Figure 1 A flow chart of the fish feeding control method provided by the present invention;

[0040] Figure 2 A schematic diagram of a flow chart of a fish feeding control method provided by the present invention;

[0041] Figure 3 A schematic diagram of the structure of the fish feeding control device provided by the present invention;

[0042] Figure 4 This is a structural schematic diagram of the fish feeding control system provided by the present invention;

[0043] Figure 5 A schematic diagram of the physical structure of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] Figure 1 The flow chart of the fish feeding control method provided by the present invention is as follows: Figure 1 As shown, an embodiment of the present invention provides a fish feeding control method, comprising:

[0046] Step S1, collecting a video of the target fish school when the target fish school is feeding;

[0047] Step S2, determining the dynamic characteristics and static characteristics of the fish school's feeding behavior based on the target fish school video; wherein the dynamic characteristics are the degree of confusion of the overall movement direction of the target fish school; and the static characteristics are the texture characteristics of the target fish school's feeding video image;

[0048] Step S3, determining the feeding behavior state of the school of fish according to the dynamic characteristics and the static characteristics; wherein the feeding behavior state of the school of fish includes: a satiation state and a hunger state;

[0049] Step S4: If it is determined that the target fish school is in a state of fullness, feeding is stopped.

[0050] Specifically, the research on fish feeding behavior analysis based on machine vision is a non-invasive, non-interference, all-weather, objective and real-time technology. By continuously processing and analyzing the video images of fish schools through cameras, the feeding behavior of fish schools can be scientifically detected and analyzed. The analysis results are used to scientifically guide feeding machinery to feed, thereby scientifically and effectively optimizing aquaculture feeding technology and realizing refined farming.

[0051] Currently, fish farming often adopts the method of feeding fish at regular intervals and in fixed quantities. However, since factors such as fish growth conditions and the environment will cause changes in fish feeding needs, the method of feeding fish at regular intervals and in fixed quantities cannot meet actual needs, and it is easy to result in insufficient feeding or feed waste.

[0052] The fish feeding control method provided by the present invention is based on a feedback idea. On the basis of the existing timed feeding method, the feeding behavior of the fish school is studied and analyzed to determine the feeding behavior state of the fish school, thereby quantitatively determining the feeding needs of the fish school and guiding the feeding equipment to perform scientific and accurate feeding.

[0053] It can be understood that the method provided by the present invention is applicable to all swimming fish. When the present invention is actually applied, the specific type and number of fish are not limited. In addition, it can also be adaptably applied to other aquatic organisms (such as jellyfish, octopus and turtle, etc.). The specific application method can be adjusted according to actual needs, and the present invention is not limited to this.

[0054] In addition, it should be noted that the method provided by the present invention is applied to a fish farming feeding device, and the specific structure of the device and the feeding control mechanism can be adaptively adjusted according to actual needs, and the present invention is not limited to this.

[0055] In order to monitor the state of the target fish school, a camera is set to complete the collection of fish school video images. For example, a CCD (Charge Coupled Device) high-definition camera is set, and the sampled video images include the normal swimming process of the fish school, the feeding process, and the post-feeding process. In addition, in the actual application of the present invention, the specific type and installation position of the camera can be determined according to actual needs, and the present invention does not limit this.

[0056] The present invention extracts the feeding behavior characteristics of fish schools from fish school feeding videos to feedback the feeding activity loudness of fish schools. In step S1, when the target fish school is in a feeding state, a target fish school video is collected, and the target fish school video includes the feeding process of the fish school.

[0057] After acquiring the target fish school video, in step S2, the fish school feeding behavior characteristics are extracted according to the target fish school video, the dynamic characteristics are determined according to the degree of confusion of the overall movement direction of the target fish school, and the static characteristics are determined according to the texture characteristics of the target fish school feeding video image.

[0058] It should be noted that the specific methods for determining the dynamic and static features based on the target fish school video (such as mathematical calculation, neural network extraction, etc.), as well as the specific calculation methods for the degree of confusion in the overall movement direction of the target fish school and the texture features of the target fish school feeding video image can all be determined according to actual needs, and the present invention is not limited to this.

[0059] After the dynamic features and the static features are determined, in step S3, the dynamic features and the static features are identified to determine whether the feeding behavior state of the fish school is a satiation state or a hunger state, thereby determining whether the fish school needs to be fed.

[0060] After determining the feeding behavior state of the fish school, in step S4, if it is determined that the target fish school is in a state of fullness, then the feeding is stopped. On the contrary, if it is determined that the target fish school is in a state of hunger, then the feeding is continued.

[0061] It can be understood that since the method provided by the present invention is applied to the condition that the fish school is being fed, as a judgment mechanism for whether to stop feeding, when it is determined to stop feeding, the control of this method ends. If it is determined to continue feeding, the method is continuously applied to determine the feeding behavior state of the fish school until it is determined that the target fish school is in a state of fullness and feeding is stopped.

[0062] In addition, the specific implementation of the method of the present invention can be achieved through a neural network model. The specific model structure and training method can be set according to actual needs, and the present invention is not limited to this.

[0063] The common modeling methods for dynamic feature recognition of fish school images in existing machine vision technologies, such as BP (back propagation) neural network modeling, support vector machine (SVM) model or linear model, are restricted to specific environments and have great limitations in image information based on a single feature. They cannot quantify feeding behavior from multiple angles and in all directions. In addition, the feature parameters have high dimensions, slow recognition speed, poor practicality, and generally low model accuracy.

[0064] Compared with the prior art, the method provided by the present invention optimizes the extraction parameters of fish feeding behavior characteristics by combining dynamic characteristics and static characteristics, and realizes quantitative analysis of fish feeding behavior in combination with feedback control, thereby reducing the misjudgment of high-density and low-density fish schools, eliminating the influence of fish physical differences and external environmental factors on alignment, improving detection speed and accuracy, thereby effectively reducing feed costs, optimizing the living environment of fish, and also reducing energy consumption for aquaculture water treatment.

[0065] The fish feeding control method provided by the present invention determines the dynamic and static characteristics of the feeding behavior of the fish school through video image analysis when the target fish school is in the feeding state, uses the dynamic and static characteristics to scientifically detect and analyze the feeding behavior of the fish school, accurately determines the feeding needs of the fish, uses the analysis results to scientifically guide feeding, effectively improves the utilization rate of bait, and improves the breeding efficiency. Not only can the utilization rate of bait be improved and the breeding efficiency be increased, but it also helps to improve the water quality environment, and has a positive effect on improving aquaculture technology and realizing refined breeding.

[0066] Optionally, the fish feeding control method provided by the present invention determines the dynamic characteristics and static characteristics of the fish school feeding behavior according to the target fish school video, specifically including:

[0067] According to the target fish school video, determine several target fish school feeding video key frames;

[0068] Preprocess the key frames of the target fish school feeding video to determine the target fish school feeding image;

[0069] According to the target fish school feeding images, the dynamic and static characteristics of the fish school feeding are determined.

[0070] Specifically, since the video frames of the target fish school video may include the normal swimming process, feeding process and post-feeding process of the fish school, in order to ensure that the dynamic and static features of the target fish school feeding can be extracted, key frame extraction and preprocessing are performed on the target fish school video before feature extraction.

[0071] Extract the video key frames that can reflect the feeding state of the target fish school in the target fish school video, and determine a number of target fish school feeding video key frames. It can be understood that in the actual application of the present invention, the key frame extraction method and specific number can be adjusted according to actual needs, and the present invention does not limit this.

[0072] After determining several target fish school feeding video key frames, the key frames are preprocessed to determine the target fish school feeding images.

[0073] It is understandable that the brightness and water quality of the external environment of fish farming may lead to poor feature extraction effects. The use of preprocessing methods can effectively reduce noise interference in the image, provide high-quality images for subsequent feature extraction, and ensure feature extraction effects. In the actual application of the present invention, specific image preprocessing methods (such as image enhancement, brightness conversion, grayscale conversion, and image filtering, etc.) can be set according to actual needs, and the present invention does not limit this.

[0074] In addition, further, considering the impact of external light sources on video images, the present invention can also add a step of identifying image brightness and adjusting the brightness of the external environment according to the image brightness, so as to provide a suitable light source for video acquisition and ensure sampling quality.

[0075] Since there is a time relationship between video frames, the degree of confusion of the overall movement direction of the target fish school is determined according to the target fish school feeding image to obtain dynamic features. Static features are determined according to the texture features of the target fish school feeding video image. The fish feeding control method provided by the present invention collects the target fish school video and extracts the video key frames when the target fish school is feeding, and pre-processes the key frames, which can effectively reduce the number of image samples, reduce the difficulty of feature extraction, and reduce noise interference in the image, thereby improving the feature extraction effect. The dynamic and static features of the fish school feeding behavior are determined by video image analysis, and the dynamic and static features are used to scientifically detect and analyze the fish school feeding behavior, accurately determine the fish feeding needs, and use the analysis results to scientifically guide feeding, effectively improve the bait utilization rate, and improve the breeding efficiency.

[0076] Optionally, according to the fish feeding control method provided by the present invention, preprocessing the target fish school feeding video key frames to determine the target fish school feeding image specifically includes:

[0077] The target fish feeding images are determined by processing the target fish feeding video key frames through homomorphic filtering and histogram equalization.

[0078] Specifically, Figure 2 The flow chart of the fish feeding control method provided by the present invention is as follows: Figure 2 As shown, considering the characteristics of fish school images, the present invention performs image information preprocessing on the key frames of the target fish school feeding video, which is specifically divided into homomorphic filtering and histogram equalization.

[0079] Homomorphic filtering is used for the key frames of the target fish feeding video to construct the background of the breeding pond, and the background is updated in real time to reduce noise interference in the image; histogram equalization is performed to enhance the contrast of the fish school against the turbid water background, determine the target fish feeding image, and provide high-quality images for the subsequent detection of the degree of chaos of the fish movement trajectory.

[0080] Since the background water of the fish school image is relatively turbid, there is a certain interference in the analysis of the foreground fish school image behavior. At the same time, the image grayscale distribution interval is relatively concentrated, making it difficult to clearly display the image details of the fish school behavior. Therefore, the background modeling is completed through homomorphic filtering and histogram equalization methods, so that the grayscale interval of the fish school image and the background is separated or the grayscale pixel distribution is made more uniform, the contrast of the fish school image is increased, and the image details are clearer, so as to achieve the purpose of image enhancement.

[0081] The fish feeding control method provided by the present invention determines the dynamic characteristics and static characteristics of the feeding behavior of the fish school through video image analysis when the target fish school is in a feeding state, and uses the dynamic characteristics and static characteristics to scientifically detect and analyze the feeding behavior of the fish school, accurately determine the feeding needs of the fish, and use the analysis results to scientifically guide feeding, effectively improve the utilization rate of bait, and improve the breeding efficiency.

[0082] Optionally, the fish feeding control method provided by the present invention determines the dynamic characteristics and static characteristics of the fish school feeding according to the target fish school feeding image, specifically including:

[0083] According to the LK optical flow method, the motion vectors of several target fish feeding images are calculated; wherein the motion vectors include the lateral motion speed and the longitudinal motion speed of the pixel points in the image;

[0084] Based on the motion vectors, the speed and the rotation angle of each motion vector are calculated;

[0085] According to the speed and angle of each motion vector, the discrete degree of speed and angle is calculated to determine the dynamic characteristics of fish feeding;

[0086] The texture features of the target fish feeding image are extracted based on the gray-level co-occurrence matrix, and the static characteristics of the fish feeding are determined.

[0087] Specifically, Figure 2 As shown, the steps of determining the dynamic characteristics and static characteristics of a school of fish in the present invention are respectively dynamic vector feature calculation and static texture feature calculation.

[0088] The feeding behavior of fish schools is mainly characterized by irregular changes in movement direction and speed. The LK optical flow method is used to calculate the changes in optical flow vectors of consecutive image frames to obtain the movement trend of fish schools, which can avoid tracking multiple targets of fish schools and reduce the difficulty.

[0089] Since the optical flow vector data of the image alone is difficult to clearly describe the feeding behavior of the fish school, in order to intuitively characterize the dynamic characteristics of the fish school's feeding behavior, the present invention statistics the movement direction vector of the fish school in each frame by establishing an optical flow direction distribution histogram, screens and statistics the main movement trend of the fish school in each frame, and then quantifies the movement direction trend of the fish school in the frame through the directional entropy value.

[0090] In the dynamic feature calculation step, the motion vector calculated by the LK (Lucas-Kanade) optical flow method is the lateral motion speed V of each pixel in the image. x and longitudinal velocity V y .

[0091] Based on the lateral movement speed V x and longitudinal velocity V y According to the speed and angle calculation formula, the speed V and angle θ of each motion vector can be calculated.

[0092] The speed V calculation formula is:

[0093]

[0094] The formula for calculating the rotation angle θ is:

[0095]

[0096] The speed V and angle θ of each motion vector calculated are statistically analyzed and normalized to obtain the distribution probability of each motion vector in the image in terms of speed and angle. The discrete degree of speed and angle is measured by calculating its information entropy, and the dynamic characteristics of fish feeding are determined based on the discrete degree of speed and angle.

[0097] In the static feature calculation step, a frame of image is randomly selected from the target fish school feeding image to determine the gray-level co-occurrence matrix, and the texture features of the target fish school feeding image are extracted based on the gray-level co-occurrence matrix to determine the static features of the fish school feeding.

[0098] It can be understood that the texture features based on the gray-level co-occurrence matrix (GLDM) can be contrast, uniformity, correlation, variance, sum average, sum variance, sum entropy, difference variance, difference average and difference entropy, etc. In the actual application of the present invention, the type and quantity of texture features contained in the static features can be adjusted according to actual needs, and the present invention is not limited to this.

[0099] In addition, since only one frame of image is needed to extract static features, while optical flow needs to be calculated to extract dynamic features, multiple frames of images are needed for feature extraction. The method for determining the image used to extract static features can be adjusted according to actual needs (such as random extraction, extraction according to rules, etc.), and the present invention is not limited to this.

[0100] For example, the video sampling resolution is 1920*1080 and the frame rate is 15fps. The length of the dynamic feature sample (feeding state, non-feeding state) is 10 seconds, i.e., 150 frames, and the length of the static feature sample (feeding state, non-feeding state) is one frame every 1 second. When extracting features based on the target fish feeding image, the dynamic features are extracted based on 10 frames of the target fish feeding image, and the static features are extracted from the last frame of the 10 frames of the target fish feeding image.

[0101] The fish feeding control method provided by the present invention, when the target fish is in a feeding state, uses video image analysis and the LK optical flow method to calculate the discrete degree of speed and angle in the target fish feeding image to determine the dynamic features; and extracts the texture features of the target fish feeding image based on the grayscale co-occurrence matrix to determine the static features. The fish feeding behavior feature extraction parameters are optimized by combining dynamic features and static features, and the quantitative analysis of the fish feeding behavior is realized in combination with the feedback control aspect, thereby reducing the misjudgment of high-density and low-density fish, eliminating the influence of fish physical differences and external environmental factors on the alignment, and improving the detection speed and accuracy. The feeding behavior of the fish is scientifically detected and analyzed, the fish feeding needs are accurately determined, and the analysis results are used to scientifically guide feeding, effectively improve the utilization rate of bait, and improve the breeding efficiency.

[0102] Optionally, according to the fish feeding control method provided by the present invention, the texture features include: angular second-order moment, inverse disparity, contrast, entropy and correlation.

[0103] Specifically, generally speaking, 14 texture features of an image can be extracted based on the gray level co-occurrence matrix. In order to simplify the calculation, the present invention selects the five most representative features to describe the fish texture information of the image. The static characteristics of the fish behavior can be quantified through the statistical properties of the five texture features, directly providing the characteristics of distinguishing textures.

[0104] In the static feature calculation step, five features are selected to describe the texture information of the fish school in the image. The texture features in the present invention include: angular second-order moment, inverse disparity, contrast, entropy and correlation.

[0105] Among them, the Angular Second Moment (ASM) represents the grayscale distribution of the image, and the calculation formula is:

[0106]

[0107] Where i represents any pixel in the grayscale image, j represents another pixel in the grayscale image, θ represents the angle between pixel i and j and the horizontal coordinate, and d represents the distance relationship between pixel i and j.

[0108] Homogeneity (HOM) represents the local changes in image texture, and the calculation formula is:

[0109]

[0110] Contrast (CON) indicates the clarity of image texture, and the calculation formula is:

[0111]

[0112] Entropy (ENT) indicates the richness of image information and reflects the complexity of image texture. The calculation formula is:

[0113]

[0114] Correlation (COR) measures the similarity of the gray-level co-occurrence matrix in the row or column direction, reflecting the local gray-level correlation. The calculation formula is:

[0115]

[0116] In the formula, μ 1 , μ 2 , σ 1 and σ 2 They are P 1 (i) and P 1 (j) The mean and variance of

[0117] The fish feeding control method provided by the present invention determines the dynamic and static features of the feeding behavior of the target fish school through video image analysis under the feeding state, selects five texture features of angular second-order moment, inverse difference, contrast, entropy and correlation as static features, simplifies the calculation, effectively reduces the complexity of static feature extraction, and improves the extraction efficiency while ensuring that the texture information of the fish school is retained. The dynamic and static features are used to scientifically detect and analyze the feeding behavior of the fish school, accurately determine the feeding needs of the fish, and use the analysis results to scientifically guide feeding, effectively improve the utilization rate of bait, and improve the breeding efficiency.

[0118] Optionally, the fish feeding control method provided by the present invention determines the feeding behavior state of the fish school according to the dynamic characteristics and the static characteristics, specifically including:

[0119] The dynamic features and static features are fused and reduced by the dimensionality reduction algorithm to determine the feeding fusion features of the target fish school;

[0120] The target fish school feeding fusion features are input into the trained target fish school feeding state recognition model to determine the feeding behavior state of the fish school.

[0121] Specifically, after determining the dynamic features and static features, the KPCA (kernel principal component analysis) algorithm is used to perform feature fusion and data dimensionality reduction on 7 feature vectors (2 dynamic features and 5 static features), and normalization is performed to determine the target fish school's feeding fusion features (feeding intensity index), that is, to fuse the dynamic features and static features, quantify the overall feeding behavior of the fish school, and determine the feeding intensity index.

[0122] It is understandable that after determining the target fish school feeding fusion feature, it is necessary to identify the feature and determine the fish school feeding behavior state. Therefore, before performing feature recognition, it is also necessary to determine a trained target fish school feeding state recognition model.

[0123] In the present invention, the target fish feeding state recognition model is trained based on a sample data set consisting of sample fish feeding fusion features and corresponding fish state. In the practical application of the present invention, the specific structure and training method of the target fish feeding state recognition model can be determined according to actual needs, and the present invention does not limit this.

[0124] The fish feeding control method provided by the present invention determines the dynamic and static characteristics of the fish school's feeding behavior through video image analysis under the target fish school's feeding state, fuses the dynamic and static characteristics and performs data dimension reduction to determine the target fish school's feeding fusion characteristics, thereby quantifying the fish school's feeding intensity, and inputs the target fish school's feeding fusion characteristics into a trained target fish school's feeding state recognition model to determine the fish school's feeding behavior state. Through the method of feature fusion and dimension reduction, redundant information and noise can be removed, the spatial complexity of the calculation can be effectively reduced, and the accuracy of the model classification can be further improved while improving the model recognition speed, the feeding behavior of the fish school can be scientifically detected and analyzed, the fish's feeding needs can be accurately determined, and the analysis results can be used to scientifically guide feeding, effectively improve the bait utilization rate, and improve the breeding efficiency.

[0125] Optionally, the fish feeding control method provided by the present invention further includes, before the step of inputting the target fish school feeding fusion feature into the trained target fish school feeding state recognition model to determine the fish school feeding behavior state:

[0126] Based on the support vector machine classifier, a fish feeding state recognition model is determined;

[0127] The radial basis function is selected as the kernel function of the support vector machine classifier, the fish feeding state recognition model is trained, and the target fish feeding state recognition model is determined.

[0128] Specifically, before performing feature recognition, a trained target fish school feeding state recognition model needs to be determined.

[0129] Based on the support vector machines (SVM) classifier, the fish feeding state recognition model is determined, the sample data set (including the sample fish feeding fusion features and the corresponding fish state) is input into the fish feeding state recognition model, the RBF (Radial Basis Function) function training feature data is selected, and the target fish feeding state recognition model is determined.

[0130] The feeding behavior characteristics of fish schools have different feature spaces before and after feeding. The problem of detecting the feeding behavior of fish schools is essentially the problem of classifying the feature points of fish schools. After extracting the static and dynamic features of the image, the SVM classifier is used to map the feature vectors to different categories. It has a good effect in obtaining the optimal solution for feature classification and enhancing the generalization ability of the model.

[0131] Compared with commonly used classifiers, such as KNN, decision tree and random forest, the use of SVM to classify and judge the attack status of fish schools has a better effect when the number of collected samples is limited. It can obtain the real-time status of fish schools and guide the precise feeding of fish schools.

[0132] In the judgment step, the reduced dimension target fish feeding fusion features are sent to the trained target fish feeding state recognition model to classify and identify the features, thereby obtaining the classification results of the fish feeding behavior state (satisfaction state / hunger state).

[0133] The fish feeding control method provided by the present invention determines the dynamic and static features of the fish school's feeding behavior through video image analysis under the target fish school's feeding state, determines the target fish school's feeding fusion features by using the dynamic and static features, and inputs the target fish school's feeding fusion features into the target fish school's feeding state recognition model to determine the fish school's feeding behavior state. The present invention uses a support vector machine classifier to construct a fish school's feeding state recognition model, selects radial basis function training feature data, and can use an SVM classifier to map feature vectors to different categories after extracting the static and dynamic features of the image, and accurately determine the fish school's feeding behavior state category. Scientifically detect and analyze the fish school's feeding behavior, accurately determine the fish's feeding needs, use the analysis results to scientifically guide feeding, effectively improve bait utilization, and improve breeding efficiency.

[0134] Taking the actual application of the present invention as an example, the fish feeding control method provided by the present invention is explained in combination with the fish feeding control device.

[0135] Figure 3 The structure diagram of the fish feeding control device provided by the present invention is as follows: Figure 3As shown, the device includes a video image acquisition module, an intelligent light source compensation module, a feeding behavior analysis module and a feeding control module. The device is installed above the circulating water aquaculture pool as a whole, and the image acquisition module is placed in the center of the aquaculture pool, 1 meter away from the water surface; the intelligent light source compensation module is installed next to the image acquisition module, and the two are connected to the feeding behavior analysis module; the program and the image are stored in the storage device; the feeding control module is connected to the feeding behavior analysis module and the feeding machine.

[0136] Among them, the video image acquisition module: consists of a CCD high-definition camera, which is used to sample the feeding process of fish schools;

[0137] Intelligent light source compensation module: used to adjust the fill light according to the actual lighting conditions to ensure the sampling quality;

[0138] Feeding behavior analysis module: used for training and detecting the feeding behavior of fish schools using SVM classifier, analyzing and classifying the sampled images, and determining the feeding behavior status of fish schools (satiated or hungry);

[0139] Feeding control module: used to convert the analysis results into a control signal of 0 or 1, and guide the feeding control module to control the opening and closing of the feeding machine through the judgment results.

[0140] Figure 3 The device shown is suitable for the fish feeding control method provided by the present invention. After the device is powered on, the feeding control module starts feeding, and at the same time the video image acquisition module starts image acquisition. The intelligent light source compensation module automatically changes the brightness according to the external environment to provide a suitable light source for image acquisition. The feeding behavior analysis module feeds back the analysis results to the feeding control module to complete closed-loop control. Based on machine vision, quantitative analysis of fish feeding behavior and feedback feeding in recirculating aquaculture are realized.

[0141] Understandably, Figure 3 The device shown is only a specific example of the practical application of the present invention and is not intended to limit the present invention.

[0142] Figure 4 The structure diagram of the fish feeding control system provided by the present invention is as follows: Figure 4 As shown, the present invention also provides a fish feeding control system, comprising: a video acquisition unit 401, a feature determination unit 402, a state recognition unit 403 and a feeding control unit 404;

[0143] The video acquisition unit 401 is used to acquire a video of the target fish school when the target fish school is feeding;

[0144] The feature determination unit 402 is used to determine the dynamic features and static features of the fish school's feeding behavior based on the target fish school video; wherein the dynamic feature is the degree of confusion of the overall movement direction of the target fish school; and the static feature is the texture feature of the target fish school's feeding video image;

[0145] The state recognition unit 403 is used to determine the feeding behavior state of the school of fish according to the dynamic characteristics and the static characteristics; wherein the feeding behavior state of the school of fish includes: a satiation state and a hunger state;

[0146] The feeding control unit 404 is used to stop feeding if it is determined that the target fish school is in a state of fullness.

[0147] Specifically, the research on fish feeding behavior analysis based on machine vision is a non-invasive, non-interference, all-weather, objective and real-time technology. By continuously processing and analyzing the video images of fish schools through cameras, the feeding behavior of fish schools can be scientifically detected and analyzed. The analysis results are used to scientifically guide feeding machinery to feed, thereby scientifically and effectively optimizing aquaculture feeding technology and realizing refined farming.

[0148] Currently, fish farming often adopts the method of feeding fish at regular intervals and in fixed quantities. However, due to factors such as fish growth conditions and environment, which will lead to changes in fish feeding needs, the method of feeding fish at regular intervals and in fixed quantities cannot meet actual needs, and it is easy to result in insufficient feeding or feed waste.

[0149] The fish feeding control system provided by the present invention is based on the feedback idea. On the basis of the existing timed feeding system, it studies and analyzes the feeding behavior of the fish school, determines the feeding behavior state of the fish school, quantitatively determines the feeding needs of the fish school, and guides the feeding equipment to perform scientific and accurate feeding.

[0150] It can be understood that the system provided by the present invention is applicable to all swimming fish. When the present invention is actually applied, the specific types and numbers of fish schools are not limited. In addition, it can also be adaptably applied to other aquatic organisms (such as jellyfish, octopus and turtle, etc.). The specific application method can be adjusted according to actual needs, and the present invention is not limited to this.

[0151] In addition, it should be noted that the system provided by the present invention is applied to a fish farming feeding device, and the specific structure of the device and the feeding control mechanism can be adaptively adjusted according to actual needs, and the present invention is not limited to this.

[0152] In order to monitor the state of the target fish school, a camera is set to complete the collection of fish school video images. For example, a CCD (Charge Coupled Device) high-definition camera is set, and the sampled video images include the normal swimming process of the fish school, the feeding process, and the post-feeding process. In addition, in the actual application of the present invention, the specific type and installation position of the camera can be determined according to actual needs, and the present invention does not limit this.

[0153] The present invention extracts the characteristics of the feeding behavior of the fish school from the fish school feeding video to feedback the feeding activity loudness of the fish school. The video acquisition unit 401 is used to collect the target fish school video when the target fish school is in the feeding state. The target fish school video includes the feeding process of the fish school.

[0154] After acquiring the target fish school video, the feature determination unit 402 is used to extract the fish school feeding behavior characteristics based on the target fish school video, determine the dynamic characteristics according to the degree of confusion of the overall movement direction of the target fish school, and determine the static characteristics according to the texture characteristics of the target fish school feeding video image.

[0155] It should be noted that the specific methods for determining the dynamic and static features based on the target fish school video (such as mathematical calculation, neural network extraction, etc.), as well as the specific calculation methods for the degree of confusion in the overall movement direction of the target fish school and the texture features of the target fish school feeding video image can all be determined according to actual needs, and the present invention is not limited to this.

[0156] After determining the dynamic features and the static features, the state recognition unit 403 is used to recognize the dynamic features and the static features, and determine whether the feeding behavior state of the fish school is a satiation state or a hunger state, so as to determine whether the fish school needs to be fed.

[0157] After determining the feeding behavior state of the fish school, the feeding control unit 404 is used to stop feeding if it is determined that the target fish school is in a state of fullness, and to continue feeding if it is determined that the target fish school is in a state of hunger.

[0158] It can be understood that since the system provided by the present invention is applied to the condition that the fish school is being fed, as a judgment mechanism for whether to stop feeding, when it is determined to stop feeding, the control of this system ends. If it is determined to continue feeding, the system is continuously applied to determine the feeding behavior state of the fish school until it is determined that the target fish school is in a state of fullness and feeding is stopped.

[0159] In addition, the specific implementation of the system of the present invention can be achieved through a neural network model. The specific model structure and training method can be set according to actual needs, and the present invention is not limited to this.

[0160] The common modeling methods for dynamic feature recognition of fish school images in existing machine vision technologies, such as BP (back propagation) neural network modeling, support vector machine (SVM) model or linear model, are restricted to specific environments and have great limitations in image information based on a single feature. They cannot quantify feeding behavior from multiple angles and in all directions. In addition, the feature parameters have high dimensions, slow recognition speed, poor practicality, and generally low model accuracy.

[0161] Compared with the prior art, the system provided by the present invention optimizes the extraction parameters of fish feeding behavior characteristics by combining dynamic characteristics and static characteristics, and realizes quantitative analysis of fish feeding behavior in combination with feedback control, thereby reducing misjudgment of high-density and low-density fish schools, eliminating the influence of differences in fish physique and external environmental factors on alignment, improving detection speed and accuracy, thereby effectively reducing feed costs, optimizing the living environment of fish, and also reducing energy consumption for aquaculture water treatment.

[0162] The fish feeding control system provided by the present invention determines the dynamic and static characteristics of the feeding behavior of the fish school through video image analysis when the target fish school is in the feeding state, uses the dynamic and static characteristics to scientifically detect and analyze the feeding behavior of the fish school, accurately determines the feeding needs of the fish, and uses the analysis results to scientifically guide feeding, effectively improve the utilization rate of bait, and improve the breeding efficiency. It can not only improve the utilization rate of bait and increase the breeding efficiency, but also help improve the water quality environment, and has a positive effect on improving aquaculture technology and realizing refined breeding.

[0163] It should be noted that the fish feeding control system provided by the present invention is used to execute the above-mentioned fish feeding control method, and its specific implementation method is consistent with the method implementation method, which will not be repeated here.

[0164] Figure 5 A schematic diagram of the physical structure of the electronic device provided by the present invention, such as Figure 5As shown, the electronic device may include: a processor 501, a communication interface 502, a memory 503 and a communication bus 504, wherein the processor 501, the communication interface 502 and the memory 503 communicate with each other through the communication bus 504. The processor 501 may call the logic instructions in the memory 503 to execute the fish feeding control method, which includes: collecting a target fish school video under the target fish school feeding state; determining the dynamic characteristics and static characteristics of the fish school feeding behavior according to the target fish school video; wherein the dynamic characteristics are the degree of confusion of the overall movement direction of the target fish school; the static characteristics are the texture characteristics of the target fish school feeding video image; determining the fish school feeding behavior state according to the dynamic characteristics and the static characteristics; wherein the fish school feeding behavior state includes: a satiation state and a hunger state; if it is determined that the target fish school is in a satiation state, then stopping feeding.

[0165] In addition, the logic instructions in the above-mentioned memory 503 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0166] On the other hand, the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the fish feeding control method provided by the above-mentioned methods, and the method includes: collecting a video of a target fish school when the target fish school is in a feeding state; determining the dynamic characteristics and static characteristics of the fish school's feeding behavior based on the target fish school video; wherein the dynamic characteristics are the degree of confusion in the overall movement direction of the target fish school; the static characteristics are the texture characteristics of the target fish school feeding video image; determining the fish school's feeding behavior state based on the dynamic characteristics and the static characteristics; wherein the fish school's feeding behavior state includes: a satiety state and a hunger state; if it is determined that the target fish school is in a satiety state, stopping feeding.

[0167] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the above-mentioned fish feeding control methods, the method comprising: collecting a video of a target fish school when the target fish school is in a feeding state; determining the dynamic characteristics and static characteristics of the fish school's feeding behavior based on the target fish school video; wherein the dynamic characteristics are the degree of confusion in the overall movement direction of the target fish school; the static characteristics are the texture characteristics of the target fish school's feeding video image; determining the fish school's feeding behavior state based on the dynamic characteristics and the static characteristics; wherein the fish school's feeding behavior state includes: a satiety state and a hunger state; if it is determined that the target fish school is in a satiety state, stopping feeding.

[0168] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, i.e., they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative effort.

[0169] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment or some parts of the embodiment.

[0170] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for controlling fish feeding, characterized in that: include: When the target fish school is feeding, collect the video of the target fish school; Determine the dynamic characteristics and static characteristics of the feeding behavior of the fish school according to the target fish school video; wherein the dynamic characteristics are the degree of confusion of the overall movement direction of the target fish school; and the static characteristics are the texture characteristics of the target fish school feeding video image; Determine the feeding behavior state of the school of fish according to the dynamic characteristics and the static characteristics; wherein the feeding behavior state of the school of fish includes: a satiation state and a hunger state; If it is determined that the target fish school is in a state of satiation, feeding is stopped.

2. The fish feeding control method according to claim 1, characterized in that: Determining the dynamic characteristics and static characteristics of the feeding behavior of the fish school based on the target fish school video specifically includes: Determining a number of target fish school feeding video key frames according to the target fish school video; Preprocessing the target fish school feeding video key frames to determine the target fish school feeding image; According to the target fish school feeding image, dynamic characteristics and static characteristics of the fish school feeding are determined.

3. The fish feeding control method according to claim 2, characterized in that: The preprocessing of the target fish school feeding video key frames to determine the target fish school feeding image specifically includes: The target fish school feeding video key frames are processed by homomorphic filtering and histogram equalization to determine the target fish school feeding image.

4. The fish feeding control method according to claim 2 or 3, characterized in that: Determining the dynamic and static characteristics of the fish school feeding according to the target fish school feeding image specifically includes: According to the LK optical flow method, the motion vectors of several target fish school feeding images are calculated; wherein the motion vectors include the lateral motion speed and the longitudinal motion speed of the pixel points in the image; Based on the motion vectors, calculating the speed and the rotation angle of each motion vector; According to the speed and the angle of each motion vector, the discrete degree of the speed and the angle of rotation is calculated to determine the dynamic characteristics of the fish school feeding; The texture features of the target fish school feeding image are extracted based on the gray level co-occurrence matrix to determine the static features of the fish school feeding.

5. The fish feeding control method according to claim 4, characterized in that: The texture features include: angular second-order moment, inverse disparity, contrast, entropy and correlation.

6. The fish feeding control method according to claim 1, characterized in that: Determining the feeding behavior state of the school of fish according to the dynamic characteristics and the static characteristics specifically includes: The dynamic features and static features are fused and reduced in dimension by a dimensionality reduction algorithm to determine the target fish school feeding fusion features; The target fish school feeding fusion feature is input into a trained target fish school feeding state recognition model to determine the fish school feeding behavior state.

7. The fish feeding control method according to claim 6, characterized in that: Before the step of inputting the target fish school feeding fusion feature into the trained target fish school feeding state recognition model to determine the fish school feeding behavior state, the step further includes: Based on the support vector machine classifier, a fish feeding state recognition model is determined; A radial basis function is selected as a kernel function of a support vector machine classifier, the fish school feeding state recognition model is trained, and a target fish school feeding state recognition model is determined.

8. A fish feeding control system, characterized in that: include: Video acquisition unit, feature determination unit, state recognition unit and feeding control unit; The video acquisition unit is used to acquire a video of the target fish school when the target fish school is feeding; The feature determination unit is used to determine the dynamic features and static features of the fish school's feeding behavior based on the target fish school video; wherein the dynamic feature is the degree of confusion of the overall movement direction of the target fish school; and the static feature is the texture feature of the target fish school's feeding video image; The state recognition unit is used to determine the feeding behavior state of the school of fish according to the dynamic characteristics and the static characteristics; wherein the feeding behavior state of the school of fish includes: a satiation state and a hunger state; The feeding control unit is used to stop feeding if it is determined that the target fish school is in a state of fullness.

9. An electronic device, characterized in that: It comprises a memory and a processor, wherein the processor and the memory communicate with each other via a bus; the memory stores program instructions executable by the processor, and the processor calls the program instructions to execute the fish feeding control method as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the fish feeding control method according to any one of claims 1 to 7 is implemented.

Images