A method and system for error compensation of feature point recognition of inconsistent morphology of crayfish

Abstract

The present application relates to the technical field of artificial intelligence, aiming at the feature point recognition error problem caused by the morphological inconsistency of crayfish, a feature point recognition error compensation method for morphological inconsistency of crayfish is proposed. First, through the feature point error compensation model based on error trend, the law of error change with position is accurately captured, and the effective compensation of recognition error is realized, thereby improving the accuracy and stability of feature point recognition. Secondly, the real-time correction mechanism is introduced, combined with the selection of multiple compensation models and the adaptive adjustment of parameters, the recognition deviation caused by the morphological inconsistency of crayfish is corrected in time, and the timeliness and accuracy of the recognition result are ensured. The innovation of the present application lies in providing a comprehensive and efficient solution, which can be widely applied in crayfish shape recognition and related fields.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method and system for compensating for errors in the identification of inconsistent feature points in crayfish morphology. Background Technology

[0002] In the international market, crayfish and their products are mostly traded as frozen processed goods, making them an important part of the global aquatic product trade. The processing flow for quick-frozen crayfish tails includes the following steps: selection of raw crayfish, cleaning, steaming, cooling, separation of head and tail, quick-freezing, and cold storage.

[0003] Frozen crayfish with heads are difficult to store for long periods. Furthermore, crayfish heads are rich in nutrients and can be processed into various products, including fertilizer, turning waste into treasure and promoting comprehensive utilization. Therefore, removing the heads is a crucial processing step. Currently, there are no fully automated shelling and head-removing machines on the market. The vast majority of crayfish head removal is done manually, which is slow, inefficient, and unhygienic, easily leading to secondary contamination of the crayfish and affecting their quality.

[0004] Chinese invention patent application number 20211167827.8 discloses a method and system for controlling the separation of crayfish head and tail based on deep learning, including the following steps: acquiring a crayfish pose image to be tested and preprocessing the image; using a deep learning-based CNN (Convolutional Neural Network) to recognize the preprocessed crayfish pose image to obtain crayfish pose information; then, through error-proofing calculation and coordinate data compensation calculation, obtaining the optimal coordinates for crayfish head and tail separation; sending the optimal coordinates for crayfish head and tail separation to a robotic arm, which then performs head and tail separation based on these coordinates. The coordinate data compensation calculation in this application includes 2D plane compensation, 3D height compensation, duplicate data removal compensation, position compensation, optimal coordinate filtering compensation, and volume compensation. 2D plane compensation involves compensating fixed values ​​in the X and Y directions and performing height compensation by measuring the width of the crayfish. The compensation coefficient C is also a fixed value, resulting in poor accuracy and stability for feature point recognition. Summary of the Invention

[0005] The purpose of this invention is to accurately capture the pattern of error change with position and thereby effectively compensate for the recognition error, thus improving the accuracy and stability of feature point recognition. In view of the above-mentioned shortcomings of the prior art, a method and system for compensating for the recognition error of crayfish morphological inconsistency feature points is proposed.

[0006] A method for compensating for errors in identifying inconsistent morphological feature points in crayfish, characterized by comprising the following steps: Data acquisition: High-definition cameras were used to acquire images of crayfish in different postures, positions, and orientations, recording the actual and visually recognized positions of the crayfish in the X and Y directions, at different angles, and in different postures; Error trend analysis: Perform error trend analysis on the preprocessed image data and establish a feature point error compensation model; Correcting the camera recognition position: By combining the error compensation model with the camera recognition system, during the real-time recognition process, the compensation parameters are calculated based on the current position and posture of the crayfish through the error compensation model. The compensation parameters in the X and Y directions are added to the X and Y coordinates of the camera recognition position, respectively, to obtain the corrected position information. Output corrected location information: The corrected camera-identified location is output for subsequent processing and application.

[0007] Before performing feature point error trend analysis, the image undergoes a series of processing steps to reduce noise, enhance features, improve image quality, and increase analysability. The preprocessing method includes the following steps: Denoising: Removing noise from the image to ensure the accuracy of subsequent analysis; Image enhancement: Enhances image contrast and sharpness, highlighting feature points; Edge detection: Detecting edge information in an image helps in locating feature points; Scale calibration: The image is scaled according to the calibration board to ensure that the positional information of the feature points is consistent with their actual physical location. The purpose of preprocessing is to improve the quality and analyzability of the image, making subsequent feature point error trend analysis more accurate and reliable.

[0008] The above technical solution is further configured as follows: the error trend analysis method is to establish a feature point error compensation model for crayfish in different postures and directions, including 0° upright, 90° upright, 0° inclined, 180° inclined, 90° lying down, 180° lying down, 0° inclined, 270° inclined, 0° sideways, 90° sideways, 180° sideways and 270° sideways.

[0009] The above technical solution is further configured as follows: the polynomial fitting method establishes a feature point error compensation model, including the following steps: S11. Collect crayfish image data under corresponding postures and orientations, and mark the error between the real position and the camera recognition position; S12. Based on the collected data, use the polynomial fitting method to fit the law of error change with positional relationship and obtain the polynomial function. S13. The polynomial function obtained by fitting is applied to the real-time recognition process to realize the real-time correction of the X and Y coordinates of the camera recognition position, thereby compensating for the errors in the X and Y directions of the visual field.

[0010] The above technical solution is further configured as follows: the error trend analysis method is to establish a feature point error compensation model using CNN for crayfish in different postures and directions, including 180° upright, 270° upright, 90° inclined, 270° inclined, 0° lying down, 270° lying down, 90° inclined and 180° inclined.

[0011] By employing the above technical solution, data collected from crayfish at different postures and angles were analyzed. For crayfish in various postures and angles—upright (0°, 90°), tilted up (0°, 180°), lying flat (90°, 180°), tilted flat (0°, 270°), and side-lying (0°, 90°, 180°, 270°)—the error between the actual and recognized positions was analyzed. In practical applications, multinomial fitting showed the highest accuracy, hence the method was chosen. For other postures and angles, CNN was used for compensation.

[0012] The above technical solution is further configured as follows: the CNN establishes a feature point error compensation model, including the following steps: S21. Prepare crayfish image data under the corresponding posture and orientation, and mark the errors in the X and Y directions between the actual position and the camera recognition position; S22. Construct a CNN model suitable for errors in the X and Y directions. This model takes a crayfish image as input and outputs the error of the X and Y coordinates of the camera's recognition position. S23. During the model training phase, labeled data is used to train the CNN model so that it can accurately learn the features of the errors in the X and Y directions. S24. After training, the CNN model can accept crayfish images as input and output the errors in the X and Y directions to achieve real-time correction of the X and Y coordinates of the camera's recognition position.

[0013] The above technical solution is further configured as follows: static parameter optimization of the error compensation model, including the following steps: S31. Define a parameter space based on the complexity of the model and the number of parameters to be optimized; S32. Divide the dataset into a training set and a validation set, train the model on the training set, and evaluate the model performance on the validation set. S33. Use cross-validation to evaluate the performance of each parameter combination on the training set; S34. Using a grid search method, search in the parameter space to find the parameter combination that achieves optimal performance.

[0014] The above technical solution is further configured as follows: The method for obtaining the dataset in step S32 includes: Data acquisition: Obtain images of crayfish, including their location, posture, and angle, using a camera; Data annotation: Based on the actual position of the crayfish on the belt and the camera's recognition position, the error between the actual position and the camera's recognition position (error in the X and Y directions) is annotated. Data cleaning: Cleaning the collected data to remove unqualified images or incorrectly labeled data, ensuring data quality and usability; To perform cross-validation, the dataset is divided into multiple mutually exclusive subsets. One subset is selected sequentially as the validation set, while the remaining subsets are used as the training set. Multiple cross-validations allow for a more comprehensive evaluation of the model's performance.

[0015] The above technical solution is further configured such that the evaluation method for parameter combination performance includes: Data partitioning: First, the dataset is divided into several mutually exclusive subsets in a certain way, with one subset serving as the validation set and the remaining subsets serving as the training set; Model training: For each parameter combination, the model is trained using a training subset. During training, the model parameters are adjusted based on the parameter combinations. Model evaluation: The trained model is evaluated using a validation subset. The accuracy of the output results is used to measure the model's performance on the validation set. Performance metrics statistics: The performance metrics (accuracy, error magnitude) obtained from K cross-validations are statistically analyzed to obtain the average performance metric. This can reduce the randomness introduced by different data partitioning.

[0016] Select the optimal parameter combination: Based on the average performance index, select the parameter combination with the best performance as the final model parameters.

[0017] The above technical solution is further configured to include dynamic parameter adaptive adjustment of the error compensation model, including the following steps: monitoring the error magnitude or accuracy of the model in real-time applications, and adaptively adjusting the model parameters according to the changes in these indicators.

[0018] The above technical solution is further configured to include: dynamic parameter adaptive adjustment of the error compensation model, including the following steps: performance index monitoring: 1. During real-time identification, continuously monitor the model's performance metrics (accuracy, error magnitude).

[0019] 2. Regularly update performance metrics and process a certain number of images.

[0020] Set threshold: 1. Set a threshold to determine whether the model performance has reached the expected level.

[0021] 2. If the model's error exceeds a pre-set threshold over several consecutive periods, it indicates that the model's performance may be degraded and parameter adjustments are necessary.

[0022] Parameter tuning strategies: Based on the monitored changes in performance metrics, different parameter tuning strategies are adopted: 1. Adjustments when errors increase: If an increase in the error of the model is detected in real-time applications, it may be due to factors such as changes in environmental conditions or data distribution shifts. At this point, you can choose to increase the compensation parameters to make the model more sensitive to the current recognition error, so as to quickly adapt to new environments or data changes. 2. Adjustments for error reduction: If the model's error decreases in real-time applications, it may be because environmental conditions have stabilized or the model has adapted to the current data distribution. In this case, the compensation parameters can be appropriately reduced to avoid over-adjustment while maintaining the stability and performance of the model. Parameter adjustment method: 1. Linear adjustment: Based on the monitored changes in performance indicators, the parameter values ​​are adjusted linearly at a certain rate; based on the increase or decrease of error, a fixed compensation amount is increased or decreased in each cycle.

[0023] 2. Exponential Adjustment: Based on the monitored changes in performance indicators, parameter values ​​are adjusted exponentially, so that the adjustment speed increases or decreases over time. If the error increases over several consecutive cycles, the parameter adjustment speed is increased to adapt to the changes more quickly; conversely, the adjustment speed is slowed down to avoid over-adjustment.

[0024] Feedback control mechanism: Design a feedback control mechanism to automatically adjust model parameters based on actual recognition results and error conditions; if the recognition result of the model in real-time application deviates significantly from the actual position, increase the compensation parameter; conversely, if the recognition result is close to the actual position, decrease the compensation parameter.

[0025] Continuous optimization: Continuously monitor model performance and optimize and adjust parameter adjustment strategies based on actual conditions.

[0026] A crayfish morphological inconsistency feature point recognition error compensation system includes an image acquisition module, an image processing module, a data optimization module, a data compensation module, and a data correction module; wherein, The image acquisition module is used to acquire crayfish pose images, and it includes a light source, a camera, and an image sensor. The image processing module is used to receive the crayfish pose image acquired by the image acquisition module and to preprocess the image; The data optimization module is used to perform error trend analysis on the preprocessed image data and establish a feature point error compensation model. The data compensation module is used to combine the error compensation model with the camera recognition system. During the real-time recognition process, the compensation parameters are calculated by the error compensation model based on the current position and posture of the crayfish. The compensation parameters in the X and Y directions are added to the X and Y coordinates of the camera recognition position, respectively, to obtain the corrected position information. The data correction module is used to optimize static parameters and adaptively adjust dynamic parameters of the error compensation model, and to correct the camera recognition position in real time.

[0027] Beneficial effects: 1. Feature Point Error Compensation Model Based on Error Trend: This invention proposes a feature point error compensation model based on error trend by deeply analyzing the feature point recognition error caused by the inconsistency of crayfish morphology. This model can accurately capture the pattern of error change with position and thereby effectively compensate for the recognition error, thus improving the accuracy and stability of feature point recognition.

[0028] 2. Introduction of Real-Time Correction Mechanism: This invention introduces a real-time correction mechanism, combining the error compensation model with the camera recognition system to achieve real-time correction of the camera's recognition position. By dynamically adjusting the compensation parameters, it can promptly correct recognition deviations caused by inconsistencies in crayfish morphology, ensuring the timeliness and accuracy of the recognition results. This dynamic parameter adaptive adjustment mechanism gives the system better robustness and adaptability, maintaining high accuracy and stability in different scenarios and environments.

[0029] By optimizing static parameters and adaptively adjusting dynamic parameters, the error compensation model can achieve optimal performance in practical applications, thereby improving the accuracy and stability of crayfish morphology recognition.

[0030] 3. Multiple Compensation Model Selection and Optimization: This invention is not limited to a single mathematical model, but provides multiple compensation model options for different situations. By selecting and optimizing two compensation models, it can effectively compensate for feature point recognition errors for crayfish with different postures and orientations, improving the accuracy and stability of recognition. Simultaneously, a parameter adaptive adjustment mechanism is introduced to dynamically adjust the parameters of the compensation model according to the actual situation to adapt to recognition errors in different scenarios, thereby further improving the system's adaptability and accuracy. Attached Figure Description

[0031] Figure 1 This is a flowchart of the crayfish morphological inconsistency feature point recognition error compensation method described in this invention.

[0032] Figure 2 The image of the crayfish described in this invention is captured by a camera within a high-precision grid where Y=0 and X=0.

[0033] Figure 3 The image of the crayfish described in this invention is captured by a camera within a high-precision grid with Y=10mm and X=0.

[0034] Figure 4 The image of the crayfish described in this invention is captured by a camera within a high-precision grid with Y=20mm and X=0.

[0035] Figure 5 Taking the crayfish lying on its side at 90° with Y equal to 0 as an example, the error distribution diagram in the X direction is shown.

[0036] Figure 6 Based on the error trend analysis described in this invention, a curve showing the change of error with position is fitted using a polynomial fitting method. Detailed Implementation

[0037] The following are specific embodiments of the present invention, which are described in conjunction with the accompanying drawings. However, the present invention is not limited to these embodiments.

[0038] A method for compensating for errors in identifying inconsistent feature points in crayfish morphology, such as... Figure 1 As shown, it includes the following steps: Data acquisition: High-definition cameras were used to acquire images of crayfish in different postures, positions and orientations, recording the actual position and visually recognized position of the crayfish in the X direction, Y direction, different angles and postures; Error trend analysis: Perform error trend analysis on the preprocessed image data and establish a feature point error compensation model; Correcting the camera recognition position: By combining the error compensation model with the camera recognition system, during the real-time recognition process, the compensation parameters are calculated based on the current position and posture of the crayfish through the error compensation model. The compensation parameters in the X and Y directions are added to the X and Y coordinates of the camera recognition position, respectively, to obtain the corrected position information. Output corrected location information: The corrected camera-identified location is output for subsequent processing and application.

[0039] Example 1 The polynomial fitting compensation model includes the following steps: Data preprocessing: A certain number of images of crayfish (crayfish at different positions, postures, and angles on the belt) are captured using a high-definition camera. The images are preprocessed, including noise reduction and image enhancement, to improve the accuracy of subsequent feature points.

[0040] Feature point extraction: Extract the feature points of the crayfish from each image, including information such as X coordinates and Y coordinates.

[0041] Data acquisition: Record the actual location of the crayfish in each image and the location identified by the camera.

[0042] Taking a crayfish lying on its side at a 90° angle as an example: The true location is the actual position of the crayfish within the high-precision grid, while the camera-recognized position is the pre-processed position.

[0043] With the Y direction set to 0, move the crayfish along the X direction and record a set of data (the crayfish's actual position and the position recognized by the camera). With the Y direction equal to 10mm, move the crayfish along the X direction and record a set of data (the crayfish's actual position and the position recognized by the camera). With the Y direction equal to 20mm, move the crayfish along the X direction and record a set of data (the crayfish's actual position and the position recognized by the camera). This process continues until the Y-direction is equal to 280mm (the belt is 280mm long). Every 10mm interval in the Y-direction, the data on the crayfish's movement in the X-direction is recorded.

[0044] For example: when the Y direction is equal to 0, record the data of movement in the X direction (interval of 10mm): (Partial image shown) Figure 2-4 As shown.

[0045] Error trend analysis: Analyze the error between the true position and the camera recognition position of each feature point, and observe the pattern of error change with position.

[0046] Taking a crayfish lying on its side at 90° with Y equal to 0 as an example: the error distribution in the X direction is as follows Figure 5 As shown.

[0047] The X-axis represents the distance the crayfish moves in the X direction, and the Y-axis represents the error between the actual position and the recognized position.

[0048] Polynomial fitting: Based on the results of error trend analysis, a polynomial fitting method is used to fit the curve of error changing with location, such as... Figure 6 As shown, the fitted curve equation is: y = 3E-10x5 - 2E-07x4 + 7E-05x3 - 0.0092x2 + 0.5501x - 10.427 (X represents the distance the crayfish moves in the X direction when the Y direction is zero, and y represents the error.) Parameter optimization: Optimize the parameters of the fitted polynomial model, such as the order of the polynomial, to obtain the best fitting effect.

[0049] Real-time correction: In practical applications, the fitted polynomial model is applied to the position recognized by the camera to correct the recognition position in real time.

[0050] Example 2 The CNN compensation model includes the following steps: Data preprocessing: A certain number of images of crayfish (crayfish at different positions, postures, and angles on the belt) are captured using a high-definition camera. The images are preprocessed, including noise reduction and image enhancement, to improve the accuracy of subsequent feature points.

[0051] Feature point extraction: Extract the feature points of the crayfish from each image, including information such as X coordinates and Y coordinates.

[0052] Data acquisition: Record the actual location of the crayfish in each image and the location identified by the camera.

[0053] Take a crayfish lying face up at a 180° angle as an example: The true location is the actual position of the crayfish within the high-precision grid, while the camera-recognized position is the pre-processed position.

[0054] 1. With the X-direction set to 0, move the crayfish along the Y-direction and record a set of data (the crayfish's actual position and the position recognized by the camera). 2. With the X-axis value equal to 10mm, move the crayfish along the Y-axis and record a set of data (the crayfish's actual position and the position recognized by the camera). 3. With the X-axis position equal to 20mm, move the crayfish along the Y-axis and record a set of data (the crayfish's actual position and the position recognized by the camera). This process continues until the X-direction is equal to 280mm (the belt width is 280mm). Every 10mm interval in the X-direction, the actual position and identification position of the crayfish are recorded.

[0055] Image annotation: Annotate the location information of the crayfish in each image, including X coordinate, Y coordinate, pose, angle, actual position and recognition position error.

[0056] Next, a CNN model is constructed to handle errors in the X and Y directions. This model takes the crayfish image as input and outputs the errors in the X and Y coordinates of the camera's recognition position.

[0057] During the model training phase, labeled data is used to train the CNN model so that it can accurately learn the features of errors in the X and Y directions.

[0058] After training, the CNN model can accept crayfish images as input and output errors in the X and Y directions to achieve real-time correction of the X and Y coordinates of the camera's recognition position.

[0059] Example 3 Static parameter optimization: I. Defining the parameter space: 1. For the polynomial fitting method, we choose the order of the polynomial as a parameter. We set the order to a range of 1 to 10.

[0060] 2. For Convolutional Neural Network (CNN) models, we can adjust the number of layers and the number of neurons per layer. We set the number of layers to a range of 1 to 5, and the number of neurons per layer to a range of 16 to 128.

[0061] II. Cross-validation: 1. Divide the dataset into 5 mutually exclusive subsets in a certain way, with 4 subsets used as the training set and the remaining subset used as the validation set.

[0062] 2. Train the model on the training set and evaluate its performance on the validation set.

[0063] 3. Repeat the above steps multiple times, randomly dividing the training set and validation set each time to reduce the randomness caused by different data partitioning.

[0064] III. Grid Search: 1. Perform a grid search in the defined parameter space, trying every combination of parameters.

[0065] 2. For polynomial fitting methods, iterate through orders from 1 to 10; for CNN models, iterate through layers from 1 to 5 and the number of neurons from 16 to 128.

[0066] 3. For each combination of parameters, calculate the model's performance metrics (accuracy, error magnitude) on the validation set.

[0067] 4. Select the optimal parameter combination as the result of static parameter optimization.

[0068] IV. Specific parameter examples: 1. Polynomial fitting method: Order: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 2. CNN model: Number of floors: 1, 2, 3, 4, 5 Number of neurons per layer: 16, 32, 64, 128.

[0069] Example 4 Dynamic parameter adaptive adjustment: I. Performance Indicator Monitoring: 1. During real-time identification, continuously monitor the model's performance metrics (accuracy, error magnitude).

[0070] 2. Regularly update performance metrics and process a certain number of images.

[0071] II. Setting Thresholds: 1. Set a threshold to determine whether the model performance has reached the expected level.

[0072] 2. If the model's error exceeds a pre-set threshold over several consecutive periods, it indicates that the model's performance may be degraded and parameter adjustments are necessary.

[0073] III. Parameter Adjustment Strategy: Based on the monitored changes in performance metrics, different parameter adjustment strategies are adopted: 1. Adjustments when errors increase: If an increase in the error of the model is detected in real-time applications, it may be due to factors such as changes in environmental conditions or data distribution shifts.

[0074] At this point, you can choose to increase the compensation parameters to make the model more sensitive to the current recognition error, so as to quickly adapt to new environments or data changes.

[0075] 2. Adjustments for error reduction: If the model's error decreases in real-time applications, it may be because environmental conditions have stabilized or the model has adapted to the current data distribution.

[0076] In this case, the compensation parameters can be appropriately reduced to avoid over-adjustment while maintaining the stability and performance of the model.

[0077] IV. Parameter Adjustment Method: 1. Linear adjustment: Based on the monitored changes in performance indicators, the parameter values ​​are adjusted linearly at a certain rate.

[0078] For example, depending on the increase or decrease of the error, a fixed amount of compensation is increased or decreased each cycle.

[0079] 2. Index Adjustment: Based on the monitored changes in performance indicators, parameter values ​​are adjusted exponentially, so that the adjustment speed increases or decreases over time.

[0080] For example, if the error increases over several consecutive periods, the adjustment speed of the parameters should be increased to adapt to the changes more quickly; conversely, the adjustment speed should be slowed down to avoid over-adjustment.

[0081] V. Feedback Control Mechanism: The design incorporates a feedback control mechanism to automatically adjust model parameters based on actual recognition results and error conditions.

[0082] For example, if the model's recognition result in a real-time application deviates significantly from the actual location, the compensation parameter should be increased; conversely, if the recognition result is close to the actual location, the compensation parameter should be decreased.

[0083] VI. Continuous Optimization: Continuously monitor model performance and optimize and adjust parameter adjustment strategies based on actual conditions.

[0084] The model's performance can be reviewed periodically, and parameters and adjustment strategies can be implemented based on new data and environmental changes to maintain the model's accuracy and stability.

[0085] For any points not covered above, existing technologies shall apply.

[0086] Although specific embodiments of the present invention have been described in detail by way of examples, those skilled in the art should understand that the above examples are for illustrative purposes only and are not intended to limit the scope of the invention. Those skilled in the art can make various modifications or additions to the described specific embodiments or use similar methods to replace them, without departing from the direction of the invention or exceeding the scope defined by the appended claims. Those skilled in the art should understand that any modifications, equivalent substitutions, improvements, etc., made to the above embodiments based on the technical essence of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for compensating for errors in identifying inconsistent morphological feature points in crayfish, characterized in that, Includes the following steps: Data acquisition: High-definition cameras were used to acquire images of crayfish in different postures, positions and orientations, recording the actual position and visually recognized position of the crayfish in the X direction, Y direction, different angles and postures; Error Trend Analysis: Error trend analysis was performed on the preprocessed image data to establish a feature point error compensation model. For crayfish in different postures and orientations, including 0° upright, 90° upright, 0° inclined, 180° inclined, 90° prone, 180° prone, 0° inclined, 270° inclined, 0° sideways, 90° sideways, 180° sideways, and 270° sideways, a multinomial fitting method was used to establish a feature point error compensation model. For crayfish in different postures and orientations, including 180° upright, 270° upright, 90° inclined, 270° inclined, 0° prone, 270° prone, 90° inclined, and 180° inclined, a CNN was used to establish a feature point error compensation model. Correcting the camera recognition position: By combining the error compensation model with the camera recognition system, during the real-time recognition process, the compensation parameters are calculated based on the current position and posture of the crayfish through the error compensation model. The compensation parameters in the X and Y directions are added to the X and Y coordinates of the camera recognition position, respectively, to obtain the corrected position information. Output corrected location information: The corrected camera-identified location is output for subsequent processing and application.

2. The method for compensating for errors in identifying inconsistent morphological feature points in crayfish according to claim 1, characterized in that: The polynomial fitting method establishes a feature point error compensation model, including the following steps: S11. Collect crayfish image data under corresponding postures and orientations, and mark the error between the real position and the camera recognition position; S12. Based on the collected data, use the polynomial fitting method to fit the law of error change with positional relationship and obtain the polynomial function. S13. The polynomial function obtained by fitting is applied to the real-time recognition process to realize the real-time correction of the X and Y coordinates of the camera recognition position, thereby compensating for the errors in the X and Y directions of the visual field.

3. The method for compensating for errors in identifying inconsistent morphological feature points of crayfish according to claim 1, characterized in that: The CNN establishes a feature point error compensation model, including the following steps: S21. Prepare crayfish image data under the corresponding posture and orientation, and mark the errors in the X and Y directions between the actual position and the camera recognition position; S22. Construct a CNN model suitable for errors in the X and Y directions. This model takes a crayfish image as input and outputs the error of the X and Y coordinates of the camera's recognition position. S23. During the model training phase, labeled data is used to train the CNN model so that it can accurately learn the features of the errors in the X and Y directions. S24. After training, the CNN model can accept crayfish images as input and output the errors in the X and Y directions to achieve real-time correction of the X and Y coordinates of the camera's recognition position.

4. The method for compensating for errors in identifying inconsistent morphological feature points in crayfish according to any one of claims 1-3, characterized in that: Static parameter optimization of the error compensation model includes the following steps: S31. Define a parameter space based on the complexity of the model and the number of parameters to be optimized; S32. Divide the dataset into a training set and a validation set, train the model on the training set, and evaluate the model performance on the validation set. S33. Use cross-validation to evaluate the performance of each parameter combination on the training set; S34. Using a grid search method, search in the parameter space to find the parameter combination that achieves optimal performance.

5. The method for compensating for errors in identifying inconsistent morphological feature points in crayfish according to claim 4, characterized in that: The method for obtaining the dataset in step S32 includes: Data acquisition: Obtain images of crayfish, including their location, posture, and angle, using a camera; Data annotation: Based on the actual position of the crayfish on the belt and the camera's recognition position, the errors in the X and Y directions between the actual position and the camera's recognition position are annotated; Data cleaning: Cleaning the collected data to remove unqualified images or incorrectly labeled data, ensuring data quality and usability; Divide the dataset into a validation set and use cross-validation: Divide the dataset into multiple mutually exclusive subsets, select one subset in turn as the validation set, and use the remaining subsets as the training set, and perform cross-validation multiple times.

6. The method for compensating for errors in identifying inconsistent morphological feature points in crayfish according to claim 4, characterized in that: Methods for evaluating the performance of parameter combinations include: Data partitioning: First, the dataset is divided into several mutually exclusive subsets in a certain way, with one subset serving as the validation set and the remaining subsets serving as the training set; Model training: For each parameter combination, the model is trained using a training subset. During training, the model parameters are adjusted according to the parameter combination. Model evaluation: The trained model is evaluated using a validation subset, and the accuracy of the output results is used to measure the model's performance on the validation set. Performance metrics statistics: The performance metrics obtained from K cross-validations, including accuracy and error magnitude, are statistically analyzed to obtain the average performance metric. Select the optimal parameter combination: Based on the average performance index, select the parameter combination with the best performance as the final model parameters.

7. The method for compensating for the recognition error of inconsistent crayfish morphological feature points according to claim 4, characterized in that: It also includes dynamic parameter adaptive adjustment of the error compensation model, including the following steps: Performance metrics monitoring: During real-time recognition, the model's performance metrics are continuously monitored, including accuracy and error magnitude. Regularly update performance metrics and process a certain number of images; Set threshold: Set a threshold to determine whether the model performance has reached the expected level; If the model's error exceeds a preset threshold over several consecutive periods, it indicates a decline in model performance, requiring parameter adjustment. Parameter adjustment strategy: Based on the monitored changes in performance indicators, different parameter adjustment strategies are adopted: increase the compensation parameter when the error increases; decrease the compensation parameter when the error decreases; Parameter adjustment method: Linear adjustment: When the error remains constant across multiple cycles, the parameter values ​​are linearly adjusted for each cycle based on the monitored changes in performance indicators. Exponential adjustment: When the error of multiple cycles gradually increases or decreases, the parameter value is adjusted exponentially according to the changes in the monitored performance indicators, so that the adjustment speed increases or decreases over time. Feedback control mechanism: Design a feedback control mechanism to automatically adjust model parameters based on actual recognition results and error conditions.

8. A crayfish morphological inconsistency feature point recognition error compensation system, characterized in that: The method according to any one of claims 1-7 includes an image acquisition module, an image processing module, a data optimization module, a data compensation module, and a data correction module; wherein, The image acquisition module is used to acquire crayfish pose images, and it includes a light source, a camera, and an image sensor. The image processing module is used to receive the crayfish pose image acquired by the image acquisition module and to preprocess the image; The data optimization module is used to perform error trend analysis on the preprocessed image data and establish a feature point error compensation model. The data compensation module is used to combine the error compensation model with the camera recognition system. During the real-time recognition process, the compensation parameters are calculated by the error compensation model based on the current position and posture of the crayfish. The compensation parameters in the X and Y directions are added to the X and Y coordinates of the camera recognition position, respectively, to obtain the corrected position information. The data correction module is used to optimize static parameters and adaptively adjust dynamic parameters of the error compensation model, and to correct the camera recognition position in real time.

Images