Method and system for quantifying ingestion intensity of fish schools

Abstract

The invention relates to the technical field of computer vision, in particular to a fish school feeding intensity quantification method and system, and the method comprises the steps: building a crucian image collection platform, and carrying out the collection of a fish school feeding image; separating the experimental target fish body and the image background from the original data to form new fish body image data, and preprocessing the data; according to the new fish body image data, behavior characteristics of individuals and groups in the fish body image are extracted to represent information of the whole image; and fusing the extracted individual and group behavior characteristics to construct an ingestion intensity quantification formula, obtaining a fish shoal ingestion intensity quantification index, and realizing continuous and accurate quantification of the fish shoal ingestion intensity. The problems that classification is too fuzzy in fish school feeding intensity identification, and fish school feeding intensity changes cannot be accurately mastered are solved. Feature extraction and analysis are carried out from two aspects of individuals and groups, comprehensive consideration of local and global characteristics of the fish school is realized, and the real fish school ingestion intensity can be reflected.

Description

Technical Field

[0001] The present invention relates to the field of computer vision technology, and in particular to a method and system for quantifying the feeding intensity of a school of fish. Background Art

[0002] Quantification refers to expressing the current state of a target in quantitative form, which facilitates accurate grasp of the target's situation and timely adjustment and control. It is widely used in many fields. Quantification of fish feeding intensity is a basic operation for feeding demand assessment in the aquaculture industry. It not only helps to grasp the feeding needs of fish in real time, but also plays an important guiding role in achieving accurate feed feeding, reducing breeding costs and improving productivity. Traditional methods for quantifying fish feeding intensity identify it as several discrete categories and cannot accurately grasp the feeding intensity of fish. Therefore, a method is needed to quantify the feeding intensity of fish in a timely, accurate and efficient manner.

[0003] With the rapid development of computer vision technology, a method for quantifying the feeding intensity of fish schools based on computer vision has emerged. Due to its advantages of accuracy and efficiency, it has become an important means to study the quantification of feeding intensity. This method first obtains the side view image of the fish body, obtains the fish target foreground by constructing the mean background, fish image preprocessing, background difference and noise reduction, and then extracts the individual characteristics and group characteristics of the fish from it. On this basis, a quantitative formula is constructed to achieve continuous quantification of the feeding intensity of the fish school.

[0004] The current methods for quantifying the feeding intensity of fish schools generally identify the feeding intensity of fish schools into several discrete categories from a single group perspective, which cannot accurately grasp the feeding situation of fish schools. To address the above problems, this paper proposes a method for continuous quantification of feeding intensity that integrates the individual and group behavioral characteristics of fish schools. First, the side view image of the fish school is collected. The fish target foreground is obtained by constructing the mean background, preprocessing the fish image, background difference and noise reduction, and the fish school characteristics are divided into individual and group characteristics for classification extraction and analysis. On this basis, a quantitative formula is constructed. Finally, the method of integrating the individual and group behavioral characteristics of the fish school is used to achieve continuous and accurate quantification of the feeding intensity of the fish school.

[0005] In the feeding intensity quantification method that integrates the individual and group behavior characteristics of fish schools, the connected area of ​​the fish school is first divided into the aggregation area and the non-aggregation area according to the size of the area according to the threshold method, and then the behavior characteristics of the fish school are divided into individual characteristics and group characteristics. The characteristics of different classifications are analyzed, and on this basis, a feeding intensity quantification formula is constructed to achieve continuous and accurate quantification of the feeding intensity of fish schools. A feeding intensity quantification method that integrates the individual and group behavior characteristics of fish schools is proposed, which realizes the continuous and accurate quantification of the feeding intensity of fish schools, and solves the problem that the classification in the identification of fish feeding intensity is too fuzzy and the feeding intensity changes of fish schools cannot be accurately grasped.

[0006] In the process of quantifying the feeding intensity of fish schools, from which levels to extract the behavioral characteristics of fish bodies and how to achieve continuous and accurate quantification of the feeding intensity of fish schools are issues that need to be addressed urgently. Summary of the invention

[0007] In view of the above problems existing in the prior art, the present invention is proposed.

[0008] Therefore, the present invention provides a method for quantifying the feeding intensity of fish schools, which can accurately extract fish body characteristics and achieve continuous and precise quantification of the feeding intensity of fish schools, providing support for achieving precise feeding in aquaculture and improving aquaculture efficiency.

[0009] To solve the above technical problems, the present invention provides the following technical solutions: a method for quantifying the feeding intensity of fish schools, comprising: building a crucian carp image acquisition platform to collect fish school feeding images; separating the experimental target fish body and the image background from the original data to form new fish body image data, and preprocessing the data; extracting the behavioral characteristics of the fish body image at the individual and group levels based on the new fish body image data to represent the information of the entire image; fusing the extracted individual and group behavioral characteristics to construct a feeding intensity quantification formula, obtain a fish school feeding intensity quantification index, and realize continuous and accurate quantification of the fish school feeding intensity.

[0010] As a preferred solution of the method for quantifying the feeding intensity of a school of fish described in the present invention, the crucian carp image acquisition platform comprises a fish tank, an aerator, a camera, and a computer;

[0011] The camera is fixed on the side of the fish tank with a bracket, 1m away from the fish tank. The camera is connected to a computer via a network cable. The collected video data of the school of fish is transmitted to the computer for storage. Video frames are captured from the collected video data of the school of fish to obtain a side view image of the school of fish.

[0012] As a preferred solution of the method for quantifying the feeding intensity of a school of fish described in the present invention, the fish body and image data include reading the background data set of the aquarium, averaging the color of each pixel in the background image, and synthesizing the color mean into a background model, which is expressed as:

[0013]

[0014] Among them, Frame k is the kth image data in the background data set, (x i ,y i ) is the pixel coordinate of the image, and time represents the duration of the data video used for background synthesis;

[0015] The target fish image is enhanced by Fourier transform. The residual bait, feces noise and holes on the surface of the fish body are removed through threshold filtering, inversion filling and opening and closing operations to obtain the image of the connected domain of the fish school. The connected areas in the image with an area smaller than the threshold are defined as impurities and removed. Inversion filling is achieved by inverting the original binary image, retaining the holes in the fish body using the threshold, removing the background, and finally superimposing the original image with the inverted image to fill the holes in the connected areas of the fish body and obtain the image of the connected domain of the fish school.

[0016] As a preferred embodiment of the method for quantifying the feeding intensity of a school of fish described in the present invention, the individual characteristics include the average distance from the individual to the water surface, which is expressed as:

[0017]

[0018] Among them, Adfs represents the average distance from the individual to the water surface, y i represents the distance from the i-th individual to the water surface, and num is the number of fish bodies in the image;

[0019] The average distance Adbi between individuals is expressed as,

[0020]

[0021] Among them, fish i represents the i-th individual, min_dis is the distance between the two closest individuals other than the individual itself, and fishes[] is the set of all individuals appearing in the image data;

[0022] The fish body inclination is expressed as,

[0023] TD=[∑abs((y right_top -y left_bottom ) / (x right_top -x left_bottom ))] / num

[0024] Among them, (y right_top -y left_bottom )、(x right_top -x left_bottom ) are the upper right and lower left pixel points in the fish image, and the slope of the line connecting the two points describes the fish posture;

[0025] The relationship between individuals and feeding points is expressed as,

[0026] F(dist, radian)=abs(dist*sqrt|radian*2|)

[0027] Where dist represents the distance between the individual and the fixed feeding point, radian is the radian value of the inclination angle of the line connecting the individual and the fixed feeding point, and F(dist, radian) is the characteristic value calculated from the relationship between the individual and the feeding point. The characteristic value is negatively correlated with the feeding intensity. The behavior of a single individual cannot represent the trend of the group and in order to reduce the occurrence of accidental phenomena, the final data of the characteristic takes the mean value of F(dist, radian) of the group;

[0028] Rifp=[∑F(dist,radian)] / num

[0029] Among them, Rifp represents the relationship between the individual and the feeding point.

[0030] As a preferred embodiment of the method for quantifying the feeding intensity of a school of fish described in the present invention, the group characteristics include the proportion of the gathering area expressed as:

[0031] PAA=[∑area i (area i >thresh)] / ∑area i *100%

[0032] Among them, area i is the area of ​​the i-th connected domain, thresh is the adaptive threshold in the area filtering algorithm, which is used to distinguish between clustered and non-clustered areas;

[0033] The percentage of clusters PIAA is expressed as,

[0034]

[0035] where gather_area i represents the number of fish in the ith cluster area, and num is the total number of fish in the image;

[0036] The fish school dispersion index is expressed as:

[0037]

[0038] Where n is the number of Delaunay triangles, L i For each triangle perimeter, Li1, L i2 , L i3 They are the three sides of the triangle respectively. DISF is the average value of the perimeter of the triangle in the Delaunay triangulation network, which represents the dispersion of the fish school.

[0039] As a preferred embodiment of the method for quantifying the feeding intensity of a school of fish described in the present invention, the feeding intensity quantification formula is expressed as follows:

[0040] LIFFI=(1000-abs(Adfs-α*DISF-Adbi)*DAI)*TD / 10*Rifp

[0041] Among them, α is the weight coefficient of DISF, DAI is the aggregation index of fish school, DAI=PAA*PIAA, Adfs indicates that the feeding intensity is negatively correlated, DISF and Adbi are positively correlated with the feeding intensity, and since the final standardization process makes a difference in the previous results, the weight coefficients in the feeding intensity quantification formula need to be reversed;

[0042] The process of obtaining the α parameter is to optimize the parameter through the idea of ​​dichotomy. The parameter is determined when the accuracy of feeding intensity recognition reaches the maximum value and no longer changes within the specified value range.

[0043] Another object of the present invention is to provide a fish feeding intensity quantification system, which can solve the problem of insufficient accuracy and difficulty in achieving continuous monitoring in existing fish feeding intensity measurement technologies by constructing a fish feeding intensity quantification system.

[0044] As a preferred solution of the system for quantifying the feeding intensity of a school of fish described in the present invention, the collection and processing module, the data separation module, the feature extraction module, and the feeding intensity quantification module;

[0045] The acquisition and processing module builds a crucian carp image acquisition platform to collect images of fish feeding;

[0046] The data separation module separates the experimental target fish body and the image background from the original data to form new fish body image data and pre-processes the data;

[0047] The feature extraction module extracts the behavioral features of the individual and group levels in the fish image based on the new fish image data to represent the information of the entire image;

[0048] The feeding intensity quantification module integrates the extracted individual and group behavior characteristics to construct a feeding intensity quantification formula, obtains the quantitative index of fish feeding intensity, and realizes continuous and accurate quantification of fish feeding intensity.

[0049] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and wherein the processor implements the steps of any one of the methods for quantifying the feeding intensity of a school of fish when executing the computer program.

[0050] A computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of any one of the methods for quantifying the feeding intensity of a school of fish are implemented.

[0051] Beneficial effects of the present invention: This method proposes a feeding intensity quantification method that integrates the individual and group behavior characteristics of fish schools, realizes the continuous and accurate quantification of the feeding intensity of fish schools, and solves the problem that the classification in the identification of fish school feeding intensity is too fuzzy and the change of fish school feeding intensity cannot be accurately grasped. A multi-level fish school feature extraction scheme is proposed, which extracts and analyzes features from both the individual and group levels, realizes the comprehensive consideration of the local and global characteristics of fish schools, and can better reflect the real feeding intensity of fish schools. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0053] Figure 1 A schematic flow chart of a method for quantifying the feeding intensity of a school of fish provided in accordance with an embodiment of the present invention.

[0054] Figure 2 A schematic diagram of changes in the intensity of a school of fish feeding activity according to a method for quantifying the feeding intensity of a school of fish provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0055] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0056] Example 1, reference Figure 1-Figure 2 , which is the first embodiment of the present invention, provides a method for quantifying the feeding intensity of a school of fish, comprising:

[0057] The first step is to collect images of fish schools. A crucian carp image acquisition platform was built in the laboratory to collect images of fish feeding. The hardware equipment included in the platform includes: 1 fish tank, 2 aerators, 1 camera (Hikvision 3T86FWDV2-I3S, 8 million pixels, 4mm focal length), 1 computer, etc. The camera was fixed to the side of the fish tank with a bracket, 1m away from the fish tank, and the camera and computer were connected via a network cable. The collected video data of the fish school was transmitted to the computer for storage. The video frames were intercepted from the collected video data of the fish school to obtain the side view image of the fish school.

[0058] The second step is fish target segmentation. Fish target segmentation refers to separating the experimental target fish and the image background from the original data to form new fish image data. It is the premise and basis for quantifying the feeding intensity of fish schools. In order to achieve better segmentation effect, this paper preprocessed the fish school image before segmentation and removed noise from the segmentation result after segmentation.

[0059] First, we build a mean background model and perform background difference on the images in the data set to separate the target fish from the image background. Since the data collection in this experiment lasts for a long time and is repeated many times, changes in external factors such as lighting in the experimental environment will affect the segmentation effect of the image data. Therefore, in the image preprocessing stage, we choose to use a synthetic mean background to replace the real background image to optimize the difficulty of image segmentation and fish recognition. Read the background data set of the fish tank, calculate the average color of each pixel in the background image, and finally synthesize these color means into a background model. The formula for establishing the mean background model is shown in Formula 1:

[0060]

[0061] Among them, Frame k is the kth image data in the background data set, (x i ,y i ) is the pixel coordinate of the image, and time represents the duration of the data video used for background synthesis.

[0062] Secondly, the target fish image is enhanced by Fourier transform. Fourier transform refers to the decomposition of an arbitrary periodic function into the sum of an infinite number of sine functions. It uses the correspondence between frequency components and image appearance to simply implement some enhancement tasks that are difficult to express in the spatial domain in the frequency domain.

[0063] Finally, through threshold filtering, inverted color filling, and opening and closing operations, the noise such as leftover bait and feces in the image is removed, and the holes on the surface of the fish body are filled to obtain the image of the connected domain of the fish school. The basic idea of ​​threshold filtering is to define the connected areas in the image with an area less than the threshold as impurities and remove them by confirming an area threshold. Inverted color filling is to invert the original binary image, and then use the threshold to retain the holes in the fish body and remove the background. Finally, the original image and the inverted image are superimposed to fill the holes in the connected area of ​​the fish body, and finally obtain the image of the connected domain of the fish school.

[0064] The third step is to extract fish body features. Fish body feature extraction refers to extracting representative information from the fish body image to represent the entire image information, which is the basis for quantifying the feeding intensity of fish schools. This paper extracts features from the fish body target image obtained after segmentation, including 4 individual features and 3 group features.

[0065] The individual characteristics used mainly include the depth of the individual, the average distance between individuals, the individual posture, and the relationship between the individual and the feeding point.

[0066] Average distance from the individual to the water surface Adfs (Average distance from water surface)

[0067] During the feeding process, fish will frequently rise and fall in order to scramble for the scattered feed. This phenomenon is related to the fish's demand for feed intake, so the distance from the individual to the water surface is one of the indicators that can measure the fish's feeding demand in aquaculture. The calculation formula is as follows:

[0068]

[0069] Among them, Adfs represents the average distance from the individual to the water surface, y i represents the distance from the ith individual to the water surface, and num is the number of fish in the image.

[0070] (2) Average distance between individuals Adbi

[0071] The average distance between individuals is the average of the distances between an individual and its two adjacent individuals. The need to ingest feed drives the fish to swim near the feed, resulting in large-scale aggregation. The stronger the feeding intensity of the fish, the higher the degree of aggregation, and the smaller the distance between individual fish. The calculation method is as follows:

[0072]

[0073] Among them, fish i represents the i-th individual, min_dis is the distance between the two closest individuals other than the individual itself, and fishes[] is the set of all individuals appearing in the image data.

[0074] (3) Tilt Degree of fish body TD (Tilt Degree)

[0075] When a school of fish is feeding, its mouth will float up while its body is hidden under water. Therefore, the school of fish in the feeding state often presents a tilted posture. This is different from the horizontal swimming posture when not feeding, which is one of the important manifestations of distinguishing the feeding state of the school of fish. The calculation formula is as follows:

[0076] TD=[∑abs((y right_top -y left_bottom ) / (x right_top -x left_bottom ))] / num (4)

[0077] Among them, (y right_top -y left_bottom )、(x right_top -x left_bottom ) are the upper right and lower left pixels in the fish image, respectively, and the slope of the line connecting the two points describes the fish posture.

[0078] (4) Relationship between individuals and feeding points

[0079] Rifp (Relationship between individuals and feeding point)

[0080] When a school of fish is feeding, it will raise its head and swim quickly toward the feeding point. This movement pattern will cause most individuals in the school of fish to approach the feeding point and then float to the surface of the water to feed. Therefore, the relationship between individuals and the feeding point can objectively reflect the changes in the feeding intensity of the school of fish. Based on this, this paper constructs the following formula:

[0081] F(dist,radian)=abs(dist*sqrt|radian*2 |) (5)

[0082] Among them, dist represents the distance between the individual and the fixed feeding point, radian is the radian value of the inclination angle of the line connecting the individual and the fixed feeding point, and F(dist, radian) is the characteristic value calculated from the relationship between the individual and the feeding point. This value is negatively correlated with the feeding intensity, that is, the larger the F(dist, radian), the weaker the corresponding feeding intensity of the fish school. Because the behavior of a single individual cannot represent the trend of the group and to reduce the possibility of accidental phenomena, the final data of this feature takes the mean F(dist, radian) of the group.

[0083] Rifp=[∑F(dist,radian)] / num(6)

[0084] Group characteristics, the group characteristics used mainly include the proportion of aggregation area, the proportion of aggregation number and the degree of fish dispersion index.

[0085] (1) Proportion of aggregation area PAA

[0086] The proportion of the area of ​​the gathering area refers to the proportion of the area of ​​the area where multiple individuals gather in the image to the area of ​​the entire connected domain. When the fish are feeding, they will gather and block the food on a large scale. The larger the proportion of the gathering area, the higher the degree of gathering, and the stronger the feeding intensity of the fish. The calculation method of this feature is as follows:

[0087] PAA=[∑area i (area i >thresh)] / ∑area i *100% (7)

[0088] Among them, area i is the area of ​​the i-th connected domain, and thresh is the adaptive threshold in the area filtering algorithm, which is used to distinguish between clustered and non-clustered areas.

[0089] (2) Proportion of cluster numbers

[0090] PIAA (Proportion of Individuals in the Aggregation Area)

[0091] The proportion of the number of aggregates is the proportion of the number of individuals in the aggregate area in the total number of fish in the statistical image. The proportion of the number of aggregates is also an indicator used to measure the aggregation degree of fish schools. Unlike PAA, it can correct the recognition error of the area method. The aggregation index of the fish school is calculated by counting the number of aggregated individuals and the number of independent individuals, and finally the proportion of individuals in the aggregate area is obtained as one of the indicators for identifying the feeding intensity of the fish school. The calculation method of this feature is as follows:

[0092]

[0093] Among them, gather_area i represents the number of fish in the ith cluster area, and num is the total number of fish in the image.

[0094] (3) DISF (Degree Index of Scatter of Fish)

[0095] DISF (Fish Dispersion Index) is a group feature used to describe the dispersion of fish schools. It constructs a centroid point set by detecting the target centroid point, and uses the point set to construct a Delaunay triangulation network. After Delaunay triangulation, the average perimeter of the triangulation network is calculated as the dispersion index of the fish school. Fish schools in a non-feeding state basically swim in the same horizontal range and do not aggregate. When feed is fed, the fish schools begin to swim to the feeding point at different angles and swimming trajectories, and aggregation phenomena of different scales appear, that is, the triangles in the triangulation network will decrease. In the Delaunay triangle, the size of each triangle in the triangulation network reflects the distance between each vertex, that is, the dispersion of the fish school. Therefore, the dispersion can be used as a quantitative indicator of the feeding behavior of the fish school. The calculation formula of the characteristic DISF is:

[0096]

[0097] Where n is the number of Delaunay triangles, L i For each triangle perimeter, L i1 , L i2 , L i3 are the three sides of the triangle respectively, and DISF is the average value of the perimeter of the triangle in the Delaunay triangulation network, which represents the dispersion of the fish school. The larger the DISF is, the lower the dispersion is, and the corresponding feeding intensity is stronger.

[0098] The fourth step is to construct a quantitative formula for feeding intensity. By integrating the four individual characteristics with the three group characteristics, a quantitative formula for feeding intensity of fish schools is constructed to achieve continuous and accurate quantification of feeding intensity of fish schools. The formula is as follows:

[0099] LIFFI=(1000-abs(Adfs-α*DISF-Adbi)*DAI)*TD / 10*Rifp (10)

[0100] Among them, α is the weight coefficient of DISF, and the participation of 1000 and 10 is the standardization of data. Among them, DAI (Degree of Aggregation Index, DAI = PAA * PIAA) is the aggregation index of fish schools. The Adfs feeding intensity in the formula is negatively correlated, and DISF and Adbi are positively correlated with the feeding intensity. Since the final standardization process performs difference processing on the previous results, the weight coefficients in the formula need to be reversed. The process of obtaining the α parameter is: the parameters are gradually optimized through the idea of ​​dichotomy, and the parameters are determined when the feeding intensity recognition accuracy reaches the maximum and no longer changes within a certain numerical range to ensure that it is the parameter selection under the best performance.

[0101] LIFFI represents the feeding intensity of fish. The larger the LIFFI value, the greater the feeding demand of the fish and the stronger the corresponding feeding intensity.

[0102] like Figure 2 As mentioned above, the early stage is before the feed is put in, there is no feed in the fish tank, and the curve is at a low level. After a long period of hunger, the fish float to the surface of the water in a very short time to scramble for food after the feed is put in. After a period of scrambling for food, the feeding intensity begins to decline, and the curve tends to be flat and gradually decreases. As time goes on, most fish feeding activities basically end, but some individuals occasionally float up to eat. At this time, the quantitative index of feeding intensity fluctuates up and down to a certain extent. Until the fish finish feeding, there is still residual bait on the water surface, but the fish no longer feed, and the curve gradually returns to the state before feeding, indicating that the feeding activity has ended.

[0103] Embodiment 2, the second embodiment of the present invention, is different from the previous embodiment in that:

[0104] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, 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. The computer software product is stored in a storage medium, including several instructions for 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 described in 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, etc., which can store program codes.

[0105] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0106] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0107] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0108] Embodiment 3, this embodiment provides a system for quantifying the feeding intensity of a school of fish, characterized by comprising: a collection and processing module, a data separation module, a feature extraction module, and a feeding intensity quantification module;

[0109] The acquisition and processing module builds a crucian carp image acquisition platform to collect images of fish feeding;

[0110] The data separation module separates the experimental target fish body and the image background from the original data to form new fish body image data and pre-processes the data;

[0111] The feature extraction module extracts the behavioral features of the individual and group levels in the fish image according to the new fish image data to represent the information of the entire image;

[0112] The feeding intensity quantification module integrates the extracted individual and group behavior characteristics to construct a feeding intensity quantification formula, obtains the quantitative index of fish feeding intensity, and realizes continuous and accurate quantification of fish feeding intensity.

[0113] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for quantifying the feeding intensity of a school of fish, characterized in that: include, Build a crucian carp image acquisition platform to collect images of fish feeding; Separate the experimental target fish body and the image background from the original data to form new fish body image data, and pre-process the data; According to the new fish image data, behavioral characteristics at both individual and group levels are extracted from the fish image to represent the information of the entire image; The individual and group behavioral characteristics extracted are integrated to construct a feeding intensity quantification formula, obtain quantitative indicators of fish feeding intensity, and achieve continuous and accurate quantification of fish feeding intensity.

2. A method for quantifying the feeding intensity of a school of fish as claimed in claim 1, characterized in that: The crucian carp image acquisition platform includes a fish tank, an aerator, a camera, and a computer; The camera is fixed on the side of the fish tank with a bracket, 1m away from the fish tank. The camera is connected to a computer via a network cable. The collected video data of the school of fish is transmitted to the computer for storage. Video frames are captured from the collected video data of the school of fish to obtain a side view image of the school of fish.

3. A method for quantifying the feeding intensity of a school of fish as claimed in claim 1, characterized in that: The fish body and image data include reading the background data set of the aquarium, averaging the color of each pixel in the background image, and synthesizing the color mean into a background model, which is expressed as: Among them, Frame k is the kth image data in the background data set, (x i ,y i ) is the pixel coordinate of the image, and time represents the duration of the data video used for background synthesis; The target fish image is enhanced by Fourier transform. The residual bait, feces noise and holes on the surface of the fish body are removed through threshold filtering, inversion filling and opening and closing operations to obtain the image of the connected domain of the fish school. The connected areas in the image with an area smaller than the threshold are defined as impurities and removed. Inversion filling is achieved by inverting the original binary image, retaining the holes in the fish body using the threshold, removing the background, and finally superimposing the original image with the inverted image to fill the holes in the connected areas of the fish body and obtain the image of the connected domain of the fish school.

4. A method for quantifying the feeding intensity of a school of fish as claimed in claim 1, characterized in that: The individual characteristics include the average distance from the individual to the water surface, expressed as, Among them, Adfs represents the average distance from the individual to the water surface, y i represents the distance from the i-th individual to the water surface, and num is the number of fish bodies in the image; The average distance adbi between individuals is expressed as, Among them, fish i represents the i-th individual, min_dis is the distance between the two closest individuals other than the individual itself, and fishes[] is the set of all individuals appearing in the image data; The fish body inclination is expressed as, TD=[Σabs((y right_top -y left_bottom ) / (x right_top -x left_bottom ))] / num Among them, (y right_top -y left_bottom )、(x right_top -x left_bottom ) are the upper right and lower left pixel points in the fish image, and the slope of the line connecting the two points describes the fish posture; The relationship between individuals and feeding points is expressed as, F(dist,radian)=abs(dist*sqrt|radian*2|) Among them, dist represents the distance between the individual and the fixed feeding point, radian is the radian value of the inclination angle of the line connecting the individual and the fixed feeding point, and F(dist, radian) is the characteristic value calculated from the relationship between the individual and the feeding point. The characteristic value is negatively correlated with the feeding intensity. The behavior of a single individual cannot represent the trend of the group and in order to reduce the occurrence of accidental phenomena, the final data of the characteristic takes the mean value of F(dist, radian) of the group; Rifpo=[∑F(dist,radian)] / num Among them, Rifp represents the relationship between the individual and the feeding point.

5. A method for quantifying the feeding intensity of a school of fish as claimed in claim 1, characterized in that: The group characteristics include the area ratio of the gathering area, which is expressed as: PAA=[Σarea i (area i >thresh)] / Σarea i *100% Among them, area i is the area of ​​the i-th connected domain, thresh is the adaptive threshold in the area filtering algorithm, which is used to distinguish between clustered and non-clustered areas; The percentage of clusters PIAA is expressed as, where gathre_area i represents the number of fish in the ith cluster area, and num is the total number of fish in the image; The fish school dispersion index is expressed as: Where n is the number of Delaunay triangles, L i For each triangle perimeter, L i1 , L i2 , L i3 They are the three sides of the triangle respectively. DISF is the average value of the perimeter of the triangle in the Delaunay triangulation network, which represents the dispersion of the fish school.

6. A method for quantifying the feeding intensity of a school of fish as claimed in claim 1, characterized in that: The quantitative formula of feeding intensity is expressed as: LIFFI=(1000-abs(Adfs-α*DISF-Adbi)*DAI)*TD / 10*Rifp Among them, α is the weight coefficient of DISF, DAI is the aggregation index of fish school, DAI=PAA*PIAA, Adfs indicates that the feeding intensity is negatively correlated, DISF and Adbi are positively correlated with the feeding intensity, and since the final standardization process makes a difference in the previous results, the weight coefficients in the feeding intensity quantification formula need to be reversed; The process of obtaining the α parameter is to optimize the parameter through the idea of ​​dichotomy. The parameter is determined when the accuracy of feeding intensity recognition reaches the maximum value and no longer changes within the specified value range.

7. A system based on the method for quantifying the feeding intensity of a school of fish according to any one of claims 1 to 6, characterized in that: It includes a collection and processing module, a data separation module, a feature extraction module, and a feeding intensity quantification module; The acquisition and processing module builds a crucian carp image acquisition platform to collect images of fish feeding; The data separation module separates the experimental target fish body and the image background from the original data to form new fish body image data and pre-processes the data; The feature extraction module extracts the behavioral features of the individual and group levels in the fish image based on the new fish image data to represent the information of the entire image; The feeding intensity quantification module integrates the extracted individual and group behavior characteristics to construct a feeding intensity quantification formula, obtains the quantitative index of fish feeding intensity, and realizes continuous and accurate quantification of fish feeding intensity.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images