A method and system for identifying individual white shrimp penaeus vannamei based on relative position constraint of pigment spots

Abstract

The present application relates to the technical field of individual identification of Litopenaeus vannamei, and particularly relates to a method and system for individual identification of Litopenaeus vannamei based on relative position constraint of pigment spots. The method obtains an image containing the back joint of the prawn and rotates the image to be horizontal; extracts anatomical key points and performs thin plate spline nonlinear distortion correction; obtains a standardized pigment spot texture image through center cropping and scale scaling; then compares the image with reference images one by one, uses local template rotation matching and fits an overall affine transformation matrix error with the constraint of overall relative position of pigment spots, calculates a single similarity result; and finally outputs an identification result based on a multiple integral cumulative voting mechanism. The present application overcomes the problem of feature matching failure caused by local appearance change and addition of pigment spots during the growth and molting process of prawns, does not rely on single feature point description, and realizes long-term, stable and non-destructive automatic individual identification.

Description

Technical Field

[0001] This invention relates to the field of Litopenaeus vannamei individual identification technology, and in particular to a method and system for Litopenaeus vannamei individual identification based on the relative position constraints of pigment spots. Background Technology

[0002] Litopenaeus vannamei, native to the eastern Pacific coast, was introduced to my country in the late 1980s. Due to its outstanding economic value, it has become one of the highest-producing farmed crustaceans in my country. Long-term individual tracking is a core requirement in Litopenaeus vannamei breeding research.

[0003] Currently, the most commonly used individual tracking methods in the industry mainly rely on physical tagging techniques, including eye tags, color tags, and fluorescent tags. However, these methods have significant technical drawbacks: on the one hand, Litopenaeus vannamei has the biological characteristic of frequent molting, which makes physical tags very easy to fall off during molting; on the other hand, physical puncture or adhesion operations can easily cause physical damage to the shrimp, triggering stress responses and seriously affecting survival rates and the objectivity of breeding data.

[0004] To overcome the limitations of physical markers, non-destructive image recognition technology based on computer vision has been gradually introduced. The distribution of pigment spots on the body surface of Litopenaeus vannamei is unique to each individual and can theoretically serve as a natural biomarker. However, the pigment spots of Litopenaeus vannamei exhibit extremely unique dynamic characteristics: at different stages of the molting cycle, the size and color of the pigment spots change significantly under the stimulation of internal and external environments; more importantly, new pigment spots are constantly generated and randomly distributed during the shrimp's growth.

[0005] Existing deep feature matching methods (such as mainstream visual algorithms like SuperGlue and LightGlue) heavily rely on local appearance descriptors. The premise for these algorithms is that "each feature point must have a stable and unique visual representation." When these techniques are applied to Litopenaeus vannamei, due to changes in the appearance of old pigment points and the large influx of new pigment points, descriptors based on local appearance suffer from severe feature drift, making it impossible to establish correct point-to-point correspondences between images across time periods.

[0006] Therefore, existing computer vision methods have failed to effectively model this biological characteristic change in pigment spots of Litopenaeus vannamei, making it difficult to achieve long-term stable individual identification. Summary of the Invention

[0007] To address the technical problem that existing methods for tracking individual Litopenaeus vannamei shrimp fail to achieve long-term tracking due to changes and additions in the appearance of pigment spots causing local feature matching failures, this invention provides a method and system for identifying individual Litopenaeus vannamei shrimp based on the relative position constraints of pigment spots.

[0008] One method for individual identification of Litopenaeus vannamei based on the relative position constraints of pigment spots includes the following steps: S1. Obtain image data of the Litopenaeus vannamei to be identified; S2. Based on the image data, the connection between the cephalothorax and abdomen of the Litopenaeus vannamei to be identified is located and marked as the region of interest using a pre-trained rotating target detection model, and the image data corresponding to the region of interest is rotated to the horizontal direction; S3. Based on the image data obtained in step S2, the key points of the anatomical structure of the Litopenaeus vannamei to be identified are extracted through a pre-trained key point detection model. Based on the key points of the anatomical structure, the image data is nonlinearly distorted by the thin plate spline transformation algorithm to obtain the corrected image data corresponding to the region of interest. S4. Based on the image data obtained in step S3, and based on the preset image data resolution ratio and pixel threshold, perform center cropping and scale scaling to obtain a standardized pigment spot texture image; S5. The standardized pigment spot texture image is used as the target recognition image and compared one by one with multiple reference images in the pre-constructed reference image library. The relative position of the pigment spots is used as a constraint and the minimum overall affine transformation error is used as the objective function to obtain the similarity results between the target recognition image and each reference image. S6. Based on the obtained similarity results, cumulative voting and scoring are performed based on the preset Litopenaeus vannamei identification integral value to obtain the individual identification results of the Litopenaeus vannamei to be identified.

[0009] Furthermore, in step S1, the image data is acquired from a top-down angle by an image acquisition device positioned directly above the Litopenaeus vannamei to be identified, and the image data includes the dorsal features of the complete cephalothorax and abdomen of the Litopenaeus vannamei to be identified.

[0010] Furthermore, in step S2, the rotating target detection model is the YOLOv11m-obb model; the step of rotating the image data corresponding to the region of interest to the horizontal direction includes constructing the minimum bounding rectangle corresponding to the region of interest obtained by positioning, performing affine rotation with the orientation angle of the minimum bounding rectangle, and making the long side of the minimum bounding rectangle parallel to the X-axis of the image coordinate system.

[0011] Furthermore, in step S3, the key point detection model is the YOLOv11m-pose model; the key points of the anatomical structure include the midpoint of the lower edge of the region of interest, the midpoint of the upper edge of the region of interest, the left endpoint of the straight line of the lower edge of the region of interest, and the right endpoint of the straight line of the lower edge of the region of interest.

[0012] Furthermore, step S4 includes the following sub-steps: S401. Based on the image data obtained in step S3, perform center cropping by retaining the middle 80% of the image data width direction and the middle 90% of the image data height direction, and remove the edge parts of the image data that do not have the individual specificity of Litopenaeus vannamei to obtain the first image data. S402. Scale the bottom edge of the width direction of the first image data to a first preset pixel threshold, and scale the side edge in the height direction by the same proportion, then crop or supplement the black background to a second preset pixel threshold to obtain a standardized pigment spot texture image.

[0013] Furthermore, step S5 includes the following sub-steps: S501. Using a standardized pigment spot texture image as the target recognition image, obtain the preset step size data, matching template size data, and local search radius data for generating the grid, and extract multiple reference images to construct a reference image dataset based on a pre-built Litopenaeus vannamei reference image library. S502. Based on the reference image dataset, and on each reference image according to the step size data, generate uniform grid points to construct a reference point set, and extract the local reference template corresponding to each reference image with each grid point as the center; S503. Based on each local reference template, within the local search radius data corresponding to the target recognition image, perform multi-angle rotation incremental matching on each local reference template, and calculate the maximum normalized cross-correlation value between the target recognition image and each local reference template, as well as the corresponding best matching point, using a normalized cross-correlation algorithm, to construct an optimal matching point set. S504. Based on the best matching point set, and based on the preset similarity threshold, remove the best matching points with similarity lower than the similarity threshold to obtain the effective matching point set, and pair it with the reference point set to obtain the average normalized cross-correlation value between the effective matching point pairs and the effective matching point pairs. S505. Based on the valid matching point pairs, an affine transformation matrix is ​​constructed by fitting through a random sampling consensus algorithm, and the validity of the affine transformation matrix is ​​checked based on the preset absolute value threshold of the matrix determinant and the condition number threshold. S506. Extract the set of interior points of the affine transformation matrix that has passed the validity test, and calculate the interior point ratio, median projection error, and normalized cross-correlation coefficient of the overlapping region for each interior point. S507. Normalize the obtained inlier ratio, median projection error, normalized cross-correlation coefficient of overlapping region and average normalized cross-correlation value respectively, and sum them based on preset weights to obtain the single comparison similarity score between the target recognition image and each reference image. S508. Based on the numerical range of the similarity score, inlier ratio, and median projection error of each single comparison, the similarity results between the target recognition image and each reference image are obtained.

[0014] Furthermore, the similarity results include highly similar, moderately similar, low similar, and dissimilar; when the similarity score of a single comparison is greater than 0.7, the inlier ratio is greater than 0.5, and the median projection error is less than 2, it is judged as highly similar. When the similarity score of a single comparison is greater than 0.55 and less than or equal to 0.7, the inlier ratio is greater than 0.4, and the median projection error is less than 3, it is judged as moderately similar. When the similarity score in a single comparison is greater than 0.4 and less than or equal to 0.55, it is judged as low similarity; When the similarity score in a single comparison is less than or equal to 0.4, it is judged as dissimilar.

[0015] Furthermore, step S6 includes the following sub-steps: S601. Based on the preset Litopenaeus vannamei recognition product value, assign corresponding product values ​​to each similarity result; S602. Calculate the total similarity score between all reference images and target images belonging to the Litopenaeus vannamei to be identified, and the frequency of being judged as highly similar; S603. When the total similarity score is greater than the first judgment threshold, the Litopenaeus vannamei to be identified is determined as a candidate individual; when the total similarity score is between the second judgment threshold and the first judgment threshold, and the frequency of being determined as highly similar is greater than or equal to the frequency threshold, the Litopenaeus vannamei to be identified is determined as a candidate individual; when there are multiple candidate individuals that meet the judgment criteria, the individual with the highest total similarity score is selected as the candidate individual.

[0016] This invention also provides a Litopenaeus vannamei individual identification system based on the relative position constraint of pigment spots. This system is implemented based on any of the Litopenaeus vannamei individual identification methods described above based on the relative position constraint of pigment spots, and includes: The image acquisition module is used to acquire image data of the Litopenaeus vannamei to be identified; The image preprocessing module is used to locate the junction of the cephalothorax and abdomen of the Litopenaeus vannamei to be identified as the region of interest based on the image data and a pre-trained rotating target detection model, and to rotate the image data corresponding to the region of interest to the horizontal direction. The key point extraction and image correction module is used to extract the key points of the anatomical structure of the Litopenaeus vannamei to be identified based on the image data obtained in step S2, through a pre-trained key point detection model, and perform nonlinear distortion correction on the image data based on the key points of the anatomical structure through a thin plate spline transformation algorithm to obtain the corrected image data corresponding to the region of interest. The image cropping and pigment spot recognition module is used to perform center cropping and scale scaling based on the image data obtained in step S3 and a preset image data resolution ratio and pixel threshold to obtain a standardized pigment spot texture image. The image correlation matching module is used to take the standardized pigment spot texture image as the target recognition image and compare it one by one with multiple reference images in the pre-built reference image library. With the relative position of the pigment spots as a constraint and the minimum overall affine transformation error as the objective function, the similarity results between the target recognition image and each reference image are obtained respectively. The individual identification and determination module is used to accumulate voting scores based on the obtained similarity results and the preset Litopenaeus vannamei identification integral value to obtain the individual identification results of the Litopenaeus vannamei to be identified.

[0017] A computer-readable storage medium for storing a computer program that, when run on a computer, causes the computer to perform any of the above-mentioned methods for individual identification of Litopenaeus vannamei based on the relative position constraints of pigment spots.

[0018] An electronic device includes: a memory for storing a computer program; and a processor for executing the computer program to implement an individual identification method for Litopenaeus vannamei based on the relative position constraints of pigment spots, as described above.

[0019] Compared with the prior art, the present invention has the following beneficial effects: This invention uses the relative positions of all pigment spots at the junction of the cephalothorax and abdomen of Litopenaeus vannamei as geometric constraints, and performs registration by minimizing the overall affine transformation error. This mechanism utilizes only the spatial topological network of the spots, ensuring that newly added pigment spots and old spots with changed appearance do not disrupt the overall matrix fitting, achieving long-term stable recognition that is molting-free and growth-insensitive. Simultaneously, by extracting specific anatomical key points and combining them with thin plate spline (TPS) transformation, this invention effectively flattens the shrimp's dorsal region with its natural curvature, eliminating nonlinear spatial distortion caused by the randomness of shrimp movement or shooting angles, greatly improving the baseline consistency of image matching. Furthermore, based on single-frame affine matching, this invention integrates multi-dimensional indicators such as inlier ratio and median error for similarity rating, and introduces a dual-threshold cumulative voting mechanism combining total integral and high-confidence frequency, effectively filtering out misjudgments caused by poor single-image quality. This significantly improves the system's fault tolerance and final recognition accuracy in complex aquaculture scenarios, adapting to non-ideal imaging conditions in actual aquaculture. Attached Figure Description

[0020] Figure 1 This is a flowchart of an individual identification method for Litopenaeus vannamei based on the relative position constraint of pigment spots proposed in an embodiment of the present invention.

[0021] Figure 2 This is a schematic diagram of the structure of an individual identification system for Litopenaeus vannamei based on the relative position constraint of pigment spots proposed in an embodiment of the present invention.

[0022] Figure 3 This is a schematic diagram of the implementation process of the method proposed in the embodiments of the present invention.

[0023] Figure 4 This is a schematic diagram of the terminal device structure of a method for individual identification of Litopenaeus vannamei based on the relative position constraint of pigment spots proposed in an embodiment of the present invention.

[0024] Figure 5 This is a schematic diagram of a computer-readable storage medium structure for an individual identification method of Litopenaeus vannamei based on the relative position constraint of pigment spots proposed in an embodiment of the present invention.

[0025] In the diagram, 200 is the terminal device, 210 is the memory, 211 is the RAM, 212 is the cache memory, 213 is the ROM, 214 is the program / utility, 215 is the program module, 220 is the processor, 230 is the bus, 240 is the external device, 250 is the I / O interface, 260 is the network adapter, and 300 is the program product. Detailed Implementation

[0026] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings, but the scope of protection of the present invention is not limited to the following description.

[0027] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention; that is, the described embodiments are only a part of the embodiments of the invention, and not all of them. The components of the embodiments of the invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0028] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0029] Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or machine that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or machine. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or machine that includes said element.

[0030] The features and performance of the present invention will be further described in detail below with reference to embodiments.

[0031] Example 1: like Figure 1 As shown, this embodiment provides a method for individual identification of Litopenaeus vannamei based on the relative position constraint of pigment spots, including the following steps: S1. Obtain image data of the Litopenaeus vannamei to be identified; Furthermore, in step S1, the image data is acquired from a top-down angle by an image acquisition device positioned directly above the Litopenaeus vannamei to be identified, and the image data includes the dorsal features of the complete cephalothorax and abdomen of the Litopenaeus vannamei to be identified.

[0032] Specifically, an image acquisition device positioned directly above the Litopenaeus vannamei to be identified was used to photograph the shrimp from a top-down angle under fixed lighting and distance. This angle allowed for unobstructed acquisition of the complete dorsal features at the junction of the cephalothorax and abdomen of the shrimp, obtaining an image of the complete dorsal junction without physical contact. This provided a baseline data source for subsequent model localization and feature segmentation, which is most beneficial for subsequent algorithm segmentation and sampling.

[0033] S2. Based on the image data, the connection between the cephalothorax and abdomen of the Litopenaeus vannamei to be identified is located and marked as the region of interest using a pre-trained rotating target detection model, and the image data corresponding to the region of interest is rotated to the horizontal direction; Furthermore, in step S2, the rotating target detection model is the YOLOv11m-obb model; the step of rotating the image data corresponding to the region of interest to the horizontal direction includes constructing the minimum bounding rectangle corresponding to the region of interest obtained by positioning, performing affine rotation with the orientation angle of the minimum bounding rectangle, and making the long side of the minimum bounding rectangle parallel to the X-axis of the image coordinate system.

[0034] Specifically, this step locates the minimum bounding rectangle of the region of interest by rotating the target detection model and aligns its long side with the X-axis of the image coordinate system by affine rotation. This solves the technical problem of overall image angle interference caused by the random swimming posture of the shrimp underwater. Existing technologies often use manual selection or generate orthogonal horizontal rectangles without orientation angles when solving the same technical problem. The technical solution of this method eliminates the global rotation variable of the initial posture of the target and realizes spatial normalization of the input image posture, reducing the angle search range and computational cost during subsequent correlation matching.

[0035] Specifically, the above implementation principle is as follows: the YOLOv11m-obb network is trained using a pre-labeled training set, the current image data is input into the network for forward inference, and an angled rectangular box containing the boundary information of the target region is obtained. Then, the affine transformation function in the image processing library is called to rotate the image pixel matrix in the opposite direction according to the deflection angle of the rectangular box, with the center of the rectangular box as the rotation axis, so that the long side of the rectangular box is parallel to the horizontal coordinate axis.

[0036] S3. Based on the image data obtained in step S2, the key points of the anatomical structure of the Litopenaeus vannamei to be identified are extracted through a pre-trained key point detection model. Based on the key points of the anatomical structure, the image data is nonlinearly distorted by the thin plate spline transformation algorithm to obtain the corrected image data corresponding to the region of interest. Furthermore, in step S3, the key point detection model is the YOLOv11m-pose model; the key points of the anatomical structure include the midpoint of the lower edge of the region of interest, the midpoint of the upper edge of the region of interest, the left endpoint of the straight line of the lower edge of the region of interest, and the right endpoint of the straight line of the lower edge of the region of interest.

[0037] Specifically, this step utilizes a key point detection model to extract specific anatomical points and applies thin-plate spline transformation to spatially flatten the image region. This solves the technical problem of nonlinear perspective distortion in images caused by the natural curvature of the shrimp's back under different shooting postures. Existing technologies often use global affine scaling or fixed mesh cropping to solve the same technical problem. This solution maps the three-dimensional features containing natural curved surfaces to a standard two-dimensional plane, eliminating nonlinear spatial deformation and ensuring the consistency of spatial features between different sampling batches.

[0038] Specifically, the image processed in step S2 is input into a pre-trained keypoint detection model (using the YOLOv11m-pose model). The model accurately locates four anatomical keypoints in the region: the midpoint of the lower edge of the region of interest, the midpoint of the upper edge, the left endpoint of the lower edge line, and the right endpoint of the lower edge line. Since the shrimp's back has a naturally curved surface, direct comparison would produce perspective distortion. Based on these four keypoints, this step uses a Thin Plate Spline (TPS) transformation algorithm to flatten the curved surface image onto a standard rectangular plane, completing nonlinear distortion correction.

[0039] S4. Based on the image data obtained in step S3, and based on the preset image data resolution ratio and pixel threshold, perform center cropping and scale scaling to obtain a standardized pigment spot texture image; Furthermore, step S4 includes the following sub-steps: S401. Based on the image data obtained in step S3, perform center cropping by retaining the middle 80% of the image data width direction and the middle 90% of the image data height direction, and remove the edge parts of the image data that do not have the individual specificity of Litopenaeus vannamei to obtain the first image data. S402. Scale the bottom edge of the width direction of the first image data to a first preset pixel threshold, and scale the side edge in the height direction by the same proportion, then crop or supplement the black background to a second preset pixel threshold to obtain a standardized pigment spot texture image.

[0040] Specifically, this step performs center cropping on the image based on a preset ratio and scales and fills the edges to a uniform specific pixel size. This solves the technical problems of irrelevant background interference at the edges of the input image and inconsistent image feature scales between different individuals. Existing technologies often use global absolute scaling or fixed pixel size direct cropping to solve the same technical problems. This solution removes edge connection regions that lack individual specificity and unifies the pixel resolution of the final input image, avoiding interference from edge environmental noise on feature extraction and meeting the scale conditions for subsequent grid segmentation.

[0041] Specifically, the above implementation principle is as follows: calculate the center coordinates of the current image width and height, determine the upper, lower, left and right boundary positions of the target retention area based on the set scaling factor, perform array slicing and cropping, calculate the scaling ratio between the original image width and the first preset pixel threshold, apply this ratio to perform proportional scaling of the cropped image width and height using a bilinear interpolation algorithm, compare the difference between the scaled height and the second preset pixel threshold, fill the upper and lower ends of the image with pixel rows with a value of 0 or remove redundant rows.

[0042] S5. The standardized pigment spot texture image is used as the target recognition image and compared one by one with multiple reference images in the pre-constructed reference image library. The relative position of the pigment spots is used as a constraint and the minimum overall affine transformation error is used as the objective function to obtain the similarity results between the target recognition image and each reference image. Furthermore, step S5 includes the following sub-steps: S501. Using a standardized pigment spot texture image as the target recognition image, obtain the preset step size data, matching template size data, and local search radius data for generating the grid, and extract multiple reference images to construct a reference image dataset based on a pre-built Litopenaeus vannamei reference image library. S502. Based on the reference image dataset, and on each reference image according to the step size data, generate uniform grid points to construct a reference point set, and extract the local reference template corresponding to each reference image with each grid point as the center; S503. Based on each local reference template, within the local search radius data corresponding to the target recognition image, perform multi-angle rotation incremental matching on each local reference template, and calculate the maximum normalized cross-correlation value between the target recognition image and each local reference template, as well as the corresponding best matching point, using a normalized cross-correlation algorithm, to construct an optimal matching point set. S504. Based on the best matching point set, and based on the preset similarity threshold, remove the best matching points with similarity lower than the similarity threshold to obtain the effective matching point set, and pair it with the reference point set to obtain the average normalized cross-correlation value between the effective matching point pairs and the effective matching point pairs. S505. Based on the valid matching point pairs, an affine transformation matrix is ​​constructed by fitting through a random sampling consensus algorithm, and the validity of the affine transformation matrix is ​​checked based on the preset absolute value threshold of the matrix determinant and the condition number threshold. S506. Extract the set of interior points of the affine transformation matrix that has passed the validity test, and calculate the interior point ratio, median projection error, and normalized cross-correlation coefficient of the overlapping region for each interior point. S507. Normalize the obtained inlier ratio, median projection error, normalized cross-correlation coefficient of overlapping region and average normalized cross-correlation value respectively, and sum them based on preset weights to obtain the single comparison similarity score between the target recognition image and each reference image. S508. Based on the numerical range of the similarity score, inlier ratio, and median projection error of each single comparison, the similarity results between the target recognition image and each reference image are obtained.

[0043] Furthermore, the similarity results include high similarity, moderate similarity, low similarity, and dissimilarity; when the similarity score of a single comparison is greater than 0.7, the inlier ratio is greater than 0.5, and the median projection error is less than 2, it is judged as highly similar; when the similarity score of a single comparison is greater than 0.55 and less than or equal to 0.7, the inlier ratio is greater than 0.4, and the median projection error is less than 3, it is judged as moderately similar; when the similarity score of a single comparison is greater than 0.4 and less than or equal to 0.55, it is judged as low similarity; when the similarity score of a single comparison is less than or equal to 0.4, it is judged as dissimilarity.

[0044] Specifically, this step performs multi-angle region search and matching based on the constructed grid points, and calculates multi-dimensional index similarity by fitting the affine transformation matrix using a random sampling consensus algorithm. This solves the technical problem that the correspondence between original feature points becomes invalid due to local appearance changes and new additions of pigment spots during the growth and molting process of Litopenaeus vannamei. Existing technologies often use a technical solution of extracting local feature descriptors for point-to-point strong matching and corresponding point retrieval when solving the same technical problem. The technical solution of this method transfers the basic constraint of image matching from single pigment spot features to the macroscopic relative topological position space of all pigment spots, enabling the algorithm to identify newly added or deformed pigment regions as outliers and remove them, thus adapting to the characteristics of the periodic changes of shrimp pigment spots.

[0045] Specifically, the above implementation principle is as follows: A sub-matrix is ​​extracted from the reference image by dividing the coordinate grid according to the step size data. The template pixels are rotated within the search radius of the target image by the step angle. The cross-correlation coefficient of the product between the template matrix and the target matrix is ​​calculated. The coordinates with the largest coefficient value are retained as matching points. Point pairs below the cross-correlation threshold are filtered out. A random sampling consistency loop is used to input the input. Three point pairs are selected each time to calculate the radial matrix model, verify the matrix determinant characteristics, and count the number of interior points that conform to the projection law of the matrix. Based on the proportion of interior points, the median value of the coordinate space offset, and the normalization result of the image intersection, a weighted linear equation is used to calculate the comprehensive score, and the final interval assignment is determined according to the conditional branch.

[0046] S6. Based on the obtained similarity results, cumulative voting and scoring are performed based on the preset Litopenaeus vannamei identification integral value to obtain the individual identification results of the Litopenaeus vannamei to be identified.

[0047] Furthermore, step S6 includes the following sub-steps: S601. Based on the preset Litopenaeus vannamei recognition product value, assign corresponding product values ​​to each similarity result; S602. Calculate the total similarity score between all reference images and target images belonging to the Litopenaeus vannamei to be identified, and the frequency of being judged as highly similar; S603. When the total similarity score is greater than the first judgment threshold, the Litopenaeus vannamei to be identified is determined as a candidate individual; when the total similarity score is between the second judgment threshold and the first judgment threshold, and the frequency of being determined as highly similar is greater than or equal to the frequency threshold, the Litopenaeus vannamei to be identified is determined as a candidate individual; when there are multiple candidate individuals that meet the judgment criteria, the individual with the highest total similarity score is selected as the candidate individual.

[0048] Specifically, this step converts the similarity result of a single image comparison into a score and combines it with a set judgment frequency to perform individual attribution determination through multiple conditional thresholds. This solves the technical problem that a single image input may be affected by accidental environmental factors such as local shading or body deformation, leading to direct recognition errors. Existing technologies often use the highest single-frame similarity score in the dataset to directly match and output the identity result when solving the same technical problem. The technical solution of this method uses a statistical decision-making mechanism based on integral weight to converge the random error of single-frame testing, thereby improving the overall accuracy of the system's determination under multiple state interferences.

[0049] Specifically, the above implementation principle is as follows: establish an integral accumulation register, traverse the comparison record table between the target image and all reference images in the image library, read the level label represented by the current record status bit, execute the register value accumulation according to the mapping score value corresponding to the label, and at the same time set up a counter to count the number of times the highest confidence status occurs. Input the accumulated total score value and the counter value into a comparator, perform logical AND-OR judgment according to the set interval boundary value, and return a unique identifier that meets the conditions or output an unmatched signal.

[0050] Example 2

[0051] like Figure 2 As shown, in a preferred embodiment of the above embodiments, a Litopenaeus vannamei individual identification system based on the relative position constraint of pigment spots is provided. This system is implemented based on any of the Litopenaeus vannamei individual identification methods based on the relative position constraint of pigment spots described above, and includes: The image acquisition module is used to acquire image data of the Litopenaeus vannamei to be identified; The image preprocessing module is used to locate the junction of the cephalothorax and abdomen of the Litopenaeus vannamei to be identified as the region of interest based on the image data and a pre-trained rotating target detection model, and to rotate the image data corresponding to the region of interest to the horizontal direction. The key point extraction and image correction module is used to extract the key points of the anatomical structure of the Litopenaeus vannamei to be identified based on the image data obtained in step S2, through a pre-trained key point detection model, and perform nonlinear distortion correction on the image data based on the key points of the anatomical structure through a thin plate spline transformation algorithm to obtain the corrected image data corresponding to the region of interest. The image cropping and pigment spot recognition module is used to perform center cropping and scale scaling based on the image data obtained in step S3 and a preset image data resolution ratio and pixel threshold to obtain a standardized pigment spot texture image. The image correlation matching module is used to take the standardized pigment spot texture image as the target recognition image and compare it one by one with multiple reference images in the pre-built reference image library. With the relative position of the pigment spots as a constraint and the minimum overall affine transformation error as the objective function, the similarity results between the target recognition image and each reference image are obtained respectively. The individual identification and determination module is used to accumulate voting scores based on the obtained similarity results and the preset Litopenaeus vannamei identification integral value to obtain the individual identification results of the Litopenaeus vannamei to be identified.

[0052] Specifically, the implementation principle of the above embodiments is as follows: First, the image acquisition module acquires image data containing the dorsal features of the connection between the complete cephalothorax and abdomen of the Litopenaeus vannamei to be identified through an image acquisition device positioned directly above the target area and transmits it to the computing and processing device. Then, the image preprocessing module receives the image data, locates the region of interest and its minimum bounding rectangle through a pre-trained rotating target detection model, and performs affine rotation of the image based on the orientation angle of the bounding rectangle, aligning its long side coordinates to the horizontal X-axis of the image coordinate system. Next, the key point extraction and image correction module uses a pre-trained key point detection model to extract four anatomical key points on the upper and lower edges of the region of interest. The anatomical key points are then used as control point coordinates to input into the thin plate spline transformation algorithm for nonlinear pixel stretching interpolation, thereby completing the flattening correction of nonlinear image distortion. Subsequently, the image cropping and pigment spot recognition module performs a fixed aspect ratio center cropping on the corrected image to remove edge pixels, and outputs a standardized pigment spot texture image that conforms to a unified preset pixel threshold by combining proportional scaling with a background boundary supplementation mechanism. Next, the image correlation matching module extracts data from the pre-built reference image library, divides the reference image into a grid based on the set step size data and extracts local templates, performs multi-step angle rotation normalization cross-correlation matching within the set search radius of the target image to obtain corresponding matching points, uses the random sampling consensus algorithm to fit the overall affine transformation matrix based on the valid matching points, filters the in-point data based on the determinant validity test and calculates multidimensional statistical indicators, and comprehensively judges the similarity results of each image in a single comparison based on the multidimensional indicators. Finally, the individual identification and determination module obtains the similarity results, performs integration and assignment according to the set Litopenaeus vannamei identification integration rules, establishes an accumulator and frequency counter for the same candidate individual, performs conditional boundary comparison after traversing the comparison results of all images, and outputs the individual identification result of the Litopenaeus vannamei to be identified when the total score and the determination frequency meet the set dual threshold rules. Furthermore, during the entire system's operation, the data flow of each data processing module is constructed based on the topological constraints of the relative positions of all pigment points. All underlying matrix fitting and coordinate projection calculation processes do not involve morphological extraction operations on a single independent feature point. The design of the data processing link shields the interference of local feature changes caused by the growth and molting of aquatic animals.

[0053] Example 3

[0054] Based on Examples 1 and 2, there is a practical application scenario for the design of a grounding system for a mountain photovoltaic system. This scenario employs a method and system for individual identification of Litopenaeus vannamei based on the relative position constraints of pigment spots, as described in the above examples. Specifically, as follows... Figure 3 As shown, the processing flow is as follows: This method includes the following steps: Step S1: Acquire a complete image of the Litopenaeus vannamei to be identified; Specifically, images of Litopenaeus vannamei were captured using a 4000×3000 pixel camera at a distance of 20cm, clearly showing the pigment spots on their body surface. A total of 126 physically tagged Litopenaeus vannamei from 10 families were continuously tracked and photographed, with images collected on the start and end dates of culture. Of these, 8 shrimp from 2 families were cultured for 74 days, while the remaining 88 shrimp were cultured for periods ranging from 23 to 54 days.

[0055] Step S2: Use the trained rotating target detection model (YOLOv11m-obb) to locate the region of interest (ROI) at the junction of the cephalothorax and abdomen, output the minimum bounding rectangle of the region and its orientation, and rotate it to the horizontal direction; Specifically, the YOLOv11m-obb model was used to train a biometric detection model at the junction of the cephalothorax and abdomen to identify the location and orientation of this region.

[0056] Step S3: Extract four anatomical structure points (midpoint of the lower edge of ROI, midpoint of the upper edge of ROI, left endpoint of the line of the lower edge of ROI, and right endpoint of the line of the lower edge of ROI) in the region using the key point detection model (YOLOv11m-pose), and use thin plate spline (TPS) transformation to correct the nonlinear distortion caused by the shooting angle. Specifically, the YOLOv11m-pose model was used to detect four anatomical structure points, and thin-plate spline transformation was used to correct nonlinear distortion caused by different shooting angles.

[0057] Step S4: Perform center cropping on the corrected ROI image (retaining 80% of the middle area in the width direction and 90% in the height direction) to remove ROI edge parts that are not specific to individuals; then scale the bottom edge of the image to 160 pixels (first preset pixel threshold), scale the height proportionally, and then crop or supplement the black background to 32 pixels (second preset pixel threshold) to obtain a standardized pigment spot texture image. Specifically, for the corrected image, the middle region is cropped to retain 80% in the width direction and 90% in the height direction to remove non-individual-specific ROI edge portions; then the bottom edge of the image is scaled to 160 pixels, and the height is scaled proportionally, and the black background is cropped or supplemented to 32 pixels to obtain a standardized pigment spot texture image. During the initial shooting, 20 images from different angles are collected for each individual as a reference image library.

[0058] Step S5: After converting the standardized image to a grayscale image, input it into the image correlation matching module and compare it one by one with the registered reference image library to calculate the similarity score. This specifically includes the following sub-steps: S501: Matching parameters determined: Reference image and target image All images are normalized after step S4, and their size is fixed at width. pixels, height Pixels. Based on extensive experimental data, and considering the distribution characteristics of pigment spots in Litopenaeus vannamei at this standardized size, the following matching parameters were determined as optimal fixed values: grid step size. pixels, template size Pixels, local search radius Pixels. The above parameters ensure a uniform distribution of matching sampling points within the standardized ROI, fully covering the pigment spot distribution area, guaranteeing matching efficiency and maximizing the differences between different individuals; S502: Dense Template Matching and Rotation Incremental Search Step size on the reference image Generate uniform grid points N represents the total number of grid points. For each grid point... :by Centered on, with dimensions of The neighborhood as a template In the target image, with The corresponding position is the center and the radius is Within the rectangular area, perform multi-angle rotation matching on the template: rotation angle Similarity is calculated using normalized cross-correlation (NCC) after each rotation; the maximum NCC value is recorded. and its corresponding best matching position ;like Below national value If the value is not specified, then the point is marked as invalid.

[0059] Final output: Reference point set Target matching point set Matching confidence Valid mark .

[0060] S503: Affine Transform Estimation Based on RANSAC: Utilizing valid matching point pairs The affine transformation matrix is ​​fitted using the Random Sample Consensus Algorithm (RANSAC). , is represented as: The maximum number of RANSAC iterations is 5000, and the reprojection error threshold is set to 3 pixels. Next, an affine transformation validity check is performed, and the absolute value of the determinant of the linear component is calculated. The value should be greater than 0.4 (to avoid excessive area scaling); and the condition number should be greater than 0.4. (Avoid excessive distortion); if the condition is not met, the model is rejected.

[0061] S504: Interior Point Statistics and Error Calculation Let the set of interior points obtained after RANSAC filtering be... The following indicators are defined as shown in Table 1; Table 1. Definitions of Interior Point Statistics and Error Calculation

[0062] S505: Calculation of overall similarity score: Define the total similarity score in a single comparison. , is represented as: The apostrophe indicates an index normalized to [0, 1]. ; ; ; ;in, This is the average NCC value when the interior points are matched to the template.

[0063] According to the score Based on the geometric quality threshold, the comparison results were divided into four levels, as shown in Table 2. Table 2. Classification of Comparison Results

[0064] Step S6: After comparing the image of the shrimp to be identified with 20 reference images of each candidate individual in the reference library according to step S5, the most likely individual identity is output based on the cumulative voting decision of the comparison results of all images of each individual.

[0065] Specifically, it includes the following steps: S601: Voting score for single-image comparison: Each comparison result is assigned points based on its level: 4 points for high similarity, 2 points for medium similarity, 1 point for low similarity, and 0 points for dissimilarity or registration failure.

[0066] S602: Individual Cumulative Score and Judgment: For candidate individuals Statistics of all belonging to The sum of integrals obtained by comparing the reference image with the image to be identified and the number of times they are "highly similar". Judgment rule: If Then it is directly judged as ;like and Then it is determined to be ;like and If so, they are classified as "suspected individuals" (not directly judged); if If no individual is identified, that individual is excluded. When multiple individuals exist, the individual with the highest total score is selected as the final output. If no individual exists, "Unidentified" is returned.

[0067] Specifically, the results of continuous tracking of the experimental group showed that the identification accuracy rate was 97.7% for 88 shrimp collected after 23 to 54 days (with 2 errors); and the identification accuracy rate was 100% for 8 shrimp collected after 74 days. The main reason for the identification errors was that shrimp with smaller weights (<3g) had too few pigment spots in their Region of Interest (ROI) (only 3-5), lacking sufficient individual uniqueness. These results indicate that the method described in this embodiment can adapt to the increase of pigment spots during shrimp growth, is not sensitive to the molting process, and can achieve long-term stable individual identification.

[0068] Example 4

[0069] like Figure 4 As shown, this embodiment proposes a terminal device for an individual identification method for Litopenaeus vannamei based on the relative position constraint of pigment spots. The terminal device includes at least one memory, at least one processor, and a bus connecting different platform systems.

[0070] The memory may include readable media in the form of volatile memory, such as RAM 211 and / or cache memory, and may further include ROM 213.

[0071] The memory also stores a computer program that can be executed by a processor, causing the processor to perform any of the above-described methods for individual identification of Litopenaeus vannamei based on the relative position constraints of pigment spots in the embodiments of this application. The specific implementation and technical effects achieved are consistent with the embodiments described above, and some details will not be repeated here. The memory may also include a program / utility having a set (at least one) of program modules, including but not limited to: an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment.

[0072] Accordingly, the processor can execute the aforementioned computer programs, as well as executable programs / utilities.

[0073] A bus can represent one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus that uses any of the various bus structures.

[0074] The terminal device can also communicate with one or more external devices, such as keyboards, pointing devices, Bluetooth devices, etc., and with one or more devices capable of interacting with the terminal device, and / or with any device that enables the terminal device to communicate with one or more other computing devices (e.g., routers, modems, etc.). This communication can be performed through I / O interfaces. Furthermore, the terminal device can communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via a network adapter. The network adapter can communicate with other modules of the terminal device via a bus. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with the terminal device, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage platforms.

[0075] Example 5

[0076] like Figure 5 As shown in Example 1, this example proposes a computer-readable storage medium for an individual identification method for Litopenaeus vannamei based on the relative position constraint of pigment spots. The computer-readable storage medium stores instructions that, when executed by a processor, implement any of the aforementioned individual identification methods for Litopenaeus vannamei based on the relative position constraint of pigment spots. The specific implementation method and the achieved technical effects are consistent with those described in the examples above, and some details will not be repeated.

[0077] This embodiment provides a program product for implementing the above-described method, which may employ a portable compact disc read-only memory (CD-ROM) and include program code, and may run on a terminal device, such as a personal computer. However, the program product of the present invention is not limited thereto. In this embodiment, the readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, apparatus, or device. The program product may employ any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof.

[0078] Computer-readable storage media may include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable storage medium may also be any readable medium other than a readable storage medium, capable of sending, propagating, or transmitting a program for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof. Program code for performing operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on a user computing device, partially on a user device, as a standalone software package, partially on a user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing devices can be connected to user computing devices via any type of network, including local area networks (LANs) or wide area networks (WANs), or they can be connected to external computing devices (e.g., via the Internet through an Internet service provider).

[0079] This invention is described from the perspectives of its intended use, effectiveness, progress, and novelty. Its practical and progressive features meet the functional enhancement and use requirements emphasized by the Patent Law. The above description and drawings are merely preferred embodiments of this application and are not intended to limit this application. Therefore, all structures, devices, features, etc., that are similar to or identical to those of this application, i.e., all equivalent substitutions or modifications made in accordance with the scope of this patent application, shall fall within the scope of protection of this patent application.

[0080] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for individual identification of Litopenaeus vannamei based on the relative position constraint of pigment spots, characterized in that, Includes the following steps: S1. Obtain image data of the Litopenaeus vannamei to be identified; S2. Based on the image data, the connection between the cephalothorax and abdomen of the Litopenaeus vannamei to be identified is located and marked as the region of interest using a pre-trained rotating target detection model, and the image data corresponding to the region of interest is rotated to the horizontal direction; S3. Based on the image data obtained in step S2, the key points of the anatomical structure of the Litopenaeus vannamei to be identified are extracted through a pre-trained key point detection model. Based on the key points of the anatomical structure, the image data is nonlinearly distorted by the thin plate spline transformation algorithm to obtain the corrected image data corresponding to the region of interest. S4. Based on the image data obtained in step S3, and based on the preset image data resolution ratio and pixel threshold, perform center cropping and scale scaling to obtain a standardized pigment spot texture image; S5. The standardized pigment spot texture image is used as the target recognition image and compared one by one with multiple reference images in the pre-constructed reference image library. The relative position of the pigment spots is used as a constraint and the minimum overall affine transformation error is used as the objective function to obtain the similarity results between the target recognition image and each reference image. S6. Based on the obtained similarity results, cumulative voting and scoring are performed based on the preset Litopenaeus vannamei identification integral value to obtain the individual identification results of the Litopenaeus vannamei to be identified.

2. The method for individual identification of Litopenaeus vannamei based on the relative position constraint of pigment spots according to claim 1, characterized in that, In step S1, the image data is acquired from a top-down angle by an image acquisition device positioned directly above the Litopenaeus vannamei to be identified. The image data includes the dorsal features of the complete cephalothorax and abdomen of the Litopenaeus vannamei to be identified.

3. The method for individual identification of Litopenaeus vannamei based on the relative position constraint of pigment spots according to claim 1, characterized in that, In step S2, the rotating target detection model is the YOLOv11m-obb model; rotating the image data corresponding to the region of interest to the horizontal direction includes constructing the minimum bounding rectangle corresponding to the region of interest obtained by positioning, performing affine rotation with the orientation angle of the minimum bounding rectangle, and making the long side of the minimum bounding rectangle parallel to the X-axis of the image coordinate system.

4. The method for individual identification of Litopenaeus vannamei based on the relative position constraint of pigment spots according to claim 1, characterized in that, In step S3, the key point detection model is the YOLOv11m-pose model; the key points of the anatomical structure include the midpoint of the lower edge of the region of interest, the midpoint of the upper edge of the region of interest, the left endpoint of the line of the lower edge of the region of interest, and the right endpoint of the line of the lower edge of the region of interest.

5. The method for individual identification of Litopenaeus vannamei based on the relative position constraint of pigment spots according to claim 1, characterized in that, Step S4 includes the following sub-steps: S401. Based on the image data obtained in step S3, perform center cropping by retaining the middle 80% of the image data width direction and the middle 90% of the image data height direction, and remove the edge parts of the image data that do not have the individual specificity of Litopenaeus vannamei to obtain the first image data. S402. Scale the bottom edge of the width direction of the first image data to a first preset pixel threshold, and scale the side edge in the height direction by the same proportion, then crop or supplement the black background to a second preset pixel threshold to obtain a standardized pigment spot texture image.

6. The method for individual identification of Litopenaeus vannamei based on the relative position constraint of pigment spots according to claim 1, characterized in that, Step S5 includes the following sub-steps: S501. Using a standardized pigment spot texture image as the target recognition image, obtain the preset step size data, matching template size data, and local search radius data for generating the grid, and extract multiple reference images to construct a reference image dataset based on a pre-built Litopenaeus vannamei reference image library. S502. Based on the reference image dataset, and on each reference image according to the step size data, generate uniform grid points to construct a reference point set, and extract the local reference template corresponding to each reference image with each grid point as the center; S503. Based on each local reference template, within the local search radius data corresponding to the target recognition image, perform multi-angle rotation incremental matching on each local reference template, and calculate the maximum normalized cross-correlation value between the target recognition image and each local reference template, as well as the corresponding best matching point, using a normalized cross-correlation algorithm, to construct an optimal matching point set. S504. Based on the best matching point set, and based on the preset similarity threshold, remove the best matching points with similarity lower than the similarity threshold to obtain the effective matching point set, and pair it with the reference point set to obtain the average normalized cross-correlation value between the effective matching point pairs and the effective matching point pairs. S505. Based on the valid matching point pairs, an affine transformation matrix is ​​constructed by fitting through a random sampling consensus algorithm, and the validity of the affine transformation matrix is ​​checked based on the preset absolute value threshold of the matrix determinant and the condition number threshold. S506. Extract the set of interior points of the affine transformation matrix that has passed the validity test, and calculate the interior point ratio, median projection error, and normalized cross-correlation coefficient of the overlapping region for each interior point. S507. Normalize the obtained inlier ratio, median projection error, normalized cross-correlation coefficient of overlapping region and average normalized cross-correlation value respectively, and sum them based on preset weights to obtain the single comparison similarity score between the target recognition image and each reference image. S508. Based on the numerical range of the similarity score, inlier ratio, and median projection error of each single comparison, the similarity results between the target recognition image and each reference image are obtained.

7. The method for individual identification of Litopenaeus vannamei based on the relative position constraint of pigment spots according to claim 6, characterized in that, The similarity results include highly similar, moderately similar, low similar, and dissimilar. When the similarity score of a single comparison is greater than 0.7, the inlier ratio is greater than 0.5, and the median projection error is less than 2, it is judged as highly similar. When the similarity score of a single comparison is greater than 0.55 and less than or equal to 0.7, the inlier ratio is greater than 0.4, and the median projection error is less than 3, it is judged as moderately similar. When the similarity score in a single comparison is greater than 0.4 and less than or equal to 0.55, it is judged as low similarity; When the similarity score in a single comparison is less than or equal to 0.4, it is judged as dissimilar.

8. The method for individual identification of Litopenaeus vannamei based on the relative position constraint of pigment spots according to claim 7, characterized in that, Step S6 includes the following sub-steps: S601. Based on the preset Litopenaeus vannamei recognition product value, assign corresponding product values ​​to each similarity result; S602. Calculate the total similarity score between all reference images and target images belonging to the Litopenaeus vannamei to be identified, and the frequency of being judged as highly similar; S603. When the total score of the similarity result is greater than the first judgment threshold, the Litopenaeus vannamei to be identified is determined to be a candidate individual; When the total similarity score is between the second and first judgment thresholds, and the frequency of being judged as highly similar is greater than or equal to the frequency threshold, the Litopenaeus vannamei to be identified is judged as a candidate individual; when there are multiple candidate individuals that meet the judgment criteria, the individual with the highest total similarity score is selected as the candidate individual.

9. A Litopenaeus vannamei individual identification system based on the relative position constraint of pigment spots, the system being implemented based on the Litopenaeus vannamei individual identification method based on the relative position constraint of pigment spots as described in any one of claims 1-8, characterized in that, include: The image acquisition module is used to acquire image data of the Litopenaeus vannamei to be identified; The image preprocessing module is used to locate the junction of the cephalothorax and abdomen of the Litopenaeus vannamei to be identified as the region of interest based on the image data and a pre-trained rotating target detection model, and to rotate the image data corresponding to the region of interest to the horizontal direction. The key point extraction and image correction module is used to extract the key points of the anatomical structure of the Litopenaeus vannamei to be identified based on the image data obtained in step S2, through a pre-trained key point detection model, and perform nonlinear distortion correction on the image data based on the key points of the anatomical structure through a thin plate spline transformation algorithm to obtain the corrected image data corresponding to the region of interest. The image cropping and pigment spot recognition module is used to perform center cropping and scale scaling based on the image data obtained in step S3 and a preset image data resolution ratio and pixel threshold to obtain a standardized pigment spot texture image. The image correlation matching module is used to take the standardized pigment spot texture image as the target recognition image and compare it one by one with multiple reference images in the pre-built reference image library. With the relative position of the pigment spots as a constraint and the minimum overall affine transformation error as the objective function, the similarity results between the target recognition image and each reference image are obtained respectively. The individual identification and determination module is used to accumulate voting scores based on the obtained similarity results and the preset Litopenaeus vannamei identification integral value to obtain the individual identification results of the Litopenaeus vannamei to be identified.

Images