A digital twin image recognition method and system for aquatic organisms

Abstract

This application provides an image recognition method and system for digital twins of aquatic organisms, applied in the field of image recognition technology. By dynamically determining the priority biological features for recognition of aquatic organisms, it effectively solves the problems of decreased recognition accuracy and data link breakage caused by changes in body surface features during the growth process. It has the advantages of effectively solving the problem of decreased individual recognition accuracy caused by dynamic changes in body surface features of aquatic organisms during the growth process, and ensuring the continuity and accuracy of health data in the digital twin model.

Description

Technical Field

[0001] This application relates to the field of image recognition technology, and in particular to an image recognition method and system for digital twins of aquatic organisms. Background Technology

[0002] In large-scale intensive aquaculture environments, image recognition systems based on digital twin technology are typically deployed to achieve precise and automated monitoring of the health status of aquatic organisms. These systems periodically collect image data of fish populations within the aquaculture area using underwater camera arrays and transmit the collected image data to a backend processing unit to build and maintain a digital twin model of each individual fish. This model aims to record the fish's identity information and continuously track its health status, thereby achieving continuous tracking and accumulation of health data for specific individuals.

[0003] However, existing technologies face significant challenges in practical operation. Aquatic organisms, especially fish, do not maintain static biological characteristics throughout their growth cycle. For example, a fish's body color may deepen or lighten with age; the number, size, and arrangement of bony plates on its back and sides may differ significantly between juvenile and adult stages; even subtle spots and patterns can change due to growth, environmental adaptation, or minor abrasions. These dynamic feature evolutions lead to significant discrepancies between the current image features of the fish and the initial features stored in the digital twin model after the fish has reached a certain stage of growth, resulting in a substantial decrease in the accuracy of individual identification. The system frequently identifies the same fish as different individuals or fails to identify existing individuals, causing the digital twin model to continuously receive health data from the same physical individual, thus breaking the health data chain.

[0004] Due to the interruption in individual identification continuity, digital twin models cannot perform effective longitudinal data analysis. To address this issue, existing systems have attempted to introduce broader feature matching thresholds. While relaxing the matching threshold can improve the recognition rate of individuals with varying characteristics to some extent, it also increases the probability of misclassifying different individuals as the same fish, especially in situations of high fish density and high visual similarity between individuals. Once misidentification occurs, health data from different individuals can be incorrectly merged into a single digital twin model, leading to chaotic and inaccurate health analysis results. It may even confuse data from healthy fish with data from diseased fish, masking the true health problems.

[0005] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0006] In view of the shortcomings of the prior art, this application provides an image recognition method and system for digital twins of aquatic organisms, which has the advantages of effectively solving the problem of decreased individual recognition accuracy caused by dynamic changes in the body surface characteristics of aquatic organisms during growth, and ensuring the continuity and accuracy of health data in the digital twin model.

[0007] In a first aspect, a method for image recognition of digital twins of aquatic organisms, the method comprising the steps of:

[0008] S1: Acquire image data of aquatic organisms, and extract macroscopic morphological parameters of the aquatic organisms based on the image data;

[0009] S2: Determine the growth stage information of the aquatic organism based on the macroscopic morphological parameters;

[0010] S3: Based on the growth stage information, determine the priority identification biological characteristics of the aquatic organism;

[0011] S4: Based on the prioritized biological characteristics, extract the individual identification features of the aquatic organism from the image data, and match the individual identification features with the historical individual identification features stored in the digital twin model to identify the physical individual of the aquatic organism;

[0012] S5: Update the individual identification features to the digital twin model to update the feature library of the physical individual.

[0013] This application proposes an image recognition method for digital twins of aquatic organisms.

[0014] Furthermore, step S5 includes:

[0015] S6: Based on the updated feature library, analyze the changing trend of the individual identification features over time to obtain the changing trend results;

[0016] S7: Compare the changing trend results with the preset range;

[0017] S8: When the change trend result deviates from the preset range, a health warning is triggered.

[0018] This application proposes an image recognition method for digital twins of aquatic organisms. By combining information on the growth stages of aquatic organisms, the biological features used for individual identification are dynamically adjusted, thereby improving the accuracy and continuity of individual identification and ensuring that the digital twin model can continuously track the health data of the same physical individual.

[0019] Furthermore, step S1 includes:

[0020] S11: Acquire the image data of the aquatic organism from multiple different angles;

[0021] S12: Obtain the outline of the aquatic organism from multiple image data, and determine whether the outline is occluded;

[0022] S13: When the outline is occluded, the occluded outline is compensated based on the unoccluded outline of the aquatic organism and known morphological features to obtain the complete outline of the aquatic organism.

[0023] S14: Extract the macroscopic morphological parameters of the aquatic organism based on the complete outline.

[0024] This application proposes an image recognition method for digital twins of aquatic organisms, which analyzes the changing trend of individual identification features over time and compares them with a preset range, thereby enabling early warning of the health status of aquatic organisms.

[0025] Furthermore, step S2 includes:

[0026] S21: Obtain the previous macroscopic morphological parameters of the aquatic organisms before a preset time period;

[0027] S22: Calculate the individual growth rate of the aquatic organism based on the current macroscopic morphological parameters and the macroscopic morphological parameters;

[0028] S23: Determine the growth stage information of the aquatic organism based on the individual growth rate.

[0029] This application proposes an image recognition method for digital twins of aquatic organisms. By fusing multi-angle image data and using contour compensation technology, it can acquire image data and macroscopic morphological parameters of aquatic organisms more comprehensively and accurately, effectively handle occlusion, and improve data quality.

[0030] Furthermore, step S23 includes:

[0031] S231: Obtain the species information of the aquatic organism, and obtain the standard growth rate curve of the aquatic organism based on the species information;

[0032] S232: Map the individual growth rate onto the standard growth rate curve to obtain the point of overlap between the individual growth rate and the standard growth rate curve;

[0033] S233: Determine the growth stage information based on the overlapping points.

[0034] Furthermore, step S3 includes:

[0035] S31: Based on the species information of the aquatic organisms, a set of standard biological characteristics corresponding to the species of aquatic organisms at different growth stages is pre-defined;

[0036] S32: Select a set of biometric features corresponding to the growth stage of the individual from the set of standard biometric features;

[0037] S33: The biometric features in the set of biometric features are identified as the priority biometric features.

[0038] Furthermore, step S4 includes:

[0039] S41: Based on the biometric priority identification features, extract individual identification features from the image data, and map the individual identification features and the historical individual identification features into the feature metric space;

[0040] S42: Within the feature metric space, calculate the distance between the mapped individual identification feature and the historical individual identification feature;

[0041] S43: Based on the distance, determine the matching result between the individual identification feature and the historical individual identification feature to identify the physical individual of the aquatic organism.

[0042] Furthermore, step S42 includes;

[0043] S421: Based on the biometric priority identification features, set weight coefficients for different feature dimensions in the feature measurement space;

[0044] S422: Based on the weighting coefficients, calculate the distance between the mapped individual identification features and the historical individual identification features.

[0045] Furthermore, step S5 includes:

[0046] S51: Obtain the acquisition time of the individual identification features;

[0047] S52: Associate the individual identification features with the collection time;

[0048] S53: Store the associated individual identification features in chronological order;

[0049] S54: Append the new individual identification feature to the end of the time-sequentially stored features to update the feature library of the physical individual.

[0050] In a second aspect, a digital twin image recognition system for aquatic organisms, characterized in that it is used to implement the method described in any one of the above claims, the system comprising:

[0051] Acquisition module: Acquires image data of aquatic organisms and extracts macroscopic morphological parameters of the aquatic organisms based on the image data;

[0052] Judgment module: Based on the macroscopic morphological parameters, determine the growth stage information of the aquatic organism;

[0053] Determination module: Based on the growth stage information, determine the priority identification biological characteristics of the aquatic organism;

[0054] Identification module: Based on the priority biological features to be identified, extract the individual identification features of the aquatic organism from the image data, and match the individual identification features with the historical individual identification features stored in the digital twin model to identify the physical individual of the aquatic organism;

[0055] Update module: Updates the individual identification features to the digital twin model to update the feature library of the physical individual.

[0056] Beneficial effects: The image recognition method and system for digital twins of aquatic organisms proposed in this application effectively solves the problems of decreased recognition accuracy and data link breakage caused by changes in body surface characteristics during the growth process of aquatic organisms by dynamically determining the priority biological features for recognition. It has the advantages of effectively solving the problem of decreased individual recognition accuracy caused by dynamic changes in body surface characteristics of aquatic organisms during the growth process, and ensuring the continuity and accuracy of health data in the digital twin model. Attached Figure Description

[0057] Figure 1 This is a flowchart of an image recognition method for digital twins of aquatic organisms proposed in this application.

[0058] Figure 2 This is a structural diagram of an image recognition system for digital twins of aquatic organisms proposed in this application.

[0059] Figure 3 This is a framework diagram of an image recognition system for digital twins of aquatic organisms proposed in this application.

[0060] Labeling Explanation: 201. Acquisition Module; 202. Judgment Module; 203. Determination Module; 204. Identification Module; 205. Update Module. Detailed Implementation

[0061] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of the embodiments of this application described and marked in the accompanying drawings can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0062] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0063] Please refer to Figure 1 A digital twin image recognition method for aquatic organisms, the method comprising the following steps:

[0064] S1: Acquire image data of aquatic organisms and extract macroscopic morphological parameters of aquatic organisms based on the image data;

[0065] S2: Determine the growth stage information of aquatic organisms based on macroscopic morphological parameters;

[0066] S3: Based on growth stage information, determine the priority identification biological characteristics of aquatic organisms;

[0067] S4: Based on the priority identification of biological features, extract the individual identification features of aquatic organisms from the image data, and match the individual identification features with the historical individual identification features stored in the digital twin model to identify the physical individual of the aquatic organism.

[0068] S5: Update the individual identification features into the digital twin model to update the feature library of the physical individual.

[0069] Among them, macroscopic morphological parameters refer to the overall external morphological characteristics of aquatic organisms at a specific point in time, such as body length, body width, body height, body shape outline, fin span, etc., which can be automatically extracted from image data using existing image processing techniques, such as edge detection, region segmentation, skeleton extraction, etc.

[0070] Growth stage information refers to the specific developmental stage of an aquatic organism in its life cycle, including but not limited to the juvenile stage, subadult stage, and adult stage. It can be determined based on the changing trends of macroscopic morphological parameters or by comparing them with a preset growth curve. Its main purpose is to dynamically understand the physiological developmental state of aquatic organisms and provide a basis for selecting appropriate individual identification characteristics.

[0071] Priority identification biological characteristics refer to the distinctive and stable biological characteristics used for individual identification of aquatic organisms at specific growth stages.

[0072] Individual identification features refer to unique biological markers extracted from images of aquatic organisms to distinguish different individuals, including but not limited to body surface texture, spots, fin shape, and body color.

[0073] A digital twin model is a virtual model constructed in digital space that corresponds to a physical aquatic organism in real time. It can store historical image data, macroscopic morphological parameters, individual identification features, growth stage information, and health status of the physical organism. Its main purpose is to achieve long-term, continuous tracking and data management of aquatic organisms.

[0074] Historical individual identification features refer to the individual identification feature data previously collected and associated with a specific physical individual stored in the digital twin model.

[0075] In some preferred embodiments, this application is implemented as follows:

[0076] First, image data of sturgeon in the breeding pond is periodically collected using an underwater high-definition camera array. This image data is transmitted to an image processing unit, which uses existing computer vision algorithms to extract the outline of each sturgeon from the images and further calculates its macroscopic morphological parameters, such as body length, body width, body height, and body size index.

[0077] Next, the system compares these macroscopic morphological parameters with the preset sturgeon growth model or with historical macroscopic morphological parameters to calculate the growth rate, thereby determining whether the sturgeon is currently in the juvenile, sub-adult, or adult stage.

[0078] Once the growth stage information is determined, such as the sub-adult stage, the system will select the priority biological features corresponding to the sub-adult stage from the preset feature library. For example, for sub-adult sturgeon, the number, size and arrangement pattern of its dorsal bony plates may be determined as priority biological features, while the distribution of spots on the body surface of juvenile fish may be weakened.

[0079] Subsequently, the image processing unit extracts specific individual identification features from the latest sturgeon images based on these prioritized biometric features. For example, it quantifies the geometric features and relative positional relationships of the bony plates using feature point detection and descriptor extraction algorithms. These extracted individual identification features are then fed into the matching module and compared with the historical individual identification features of the sturgeon stored in the digital twin model. For example, by calculating the distance between feature vectors in the feature space, it confirms whether the sturgeon in the current image is the same physical individual tracked by the digital twin model. Once identification is successful, the latest individual identification features, along with their acquisition time, are added to the feature library of the sturgeon's digital twin model, ensuring that its feature information remains up-to-date, thereby supporting continuous monitoring and analysis of the physical individual's health status.

[0080] Through the above technical solution, this application effectively addresses the problem of decreased individual identification accuracy caused by changes in the surface biological characteristics of aquatic organisms during their growth. By dynamically judging the growth stage information of aquatic organisms and determining the priority biological characteristics for identification accordingly, the system can adapt to the morphological changes of organisms at different growth stages, thereby improving the accuracy and robustness of individual identification. This ensures that the digital twin model can continuously and stably track the same physical individual, avoiding data chain breaks caused by identification interruptions, and thus guaranteeing the continuous accumulation and analysis of health data, providing a data foundation for the refined management and early health warning of aquatic organisms.

[0081] Furthermore, step S5 includes:

[0082] S6: Based on the updated feature library, analyze the changing trend of individual identification features over time to obtain the changing trend results;

[0083] S7: Compare the trend results with the preset range;

[0084] S8: When the trend result deviates from the preset range, a health warning is triggered.

[0085] Among them, the trend of change refers to the continuous change pattern of individual identification characteristics of aquatic organisms, such as spots, colors, and shapes on their body surfaces, at different points in time. It can be obtained by time series analysis, with the aim of capturing the dynamic evolution of the health status of aquatic organisms and discovering potential abnormal signals.

[0086] The trend result refers to the quantitative or qualitative description obtained by analyzing the trend of individual identification characteristics over time, including but not limited to the rate of change, fluctuation range, and degree of deviation.

[0087] The preset range refers to the range or threshold of individual identification characteristics under normal health conditions that are pre-set based on factors such as the species, growth stage, and environmental conditions of aquatic organisms. Its purpose is to serve as a benchmark for judging whether the health status of aquatic organisms is abnormal.

[0088] In some preferred embodiments, this application is implemented as follows:

[0089] After updating the individual identification features to the digital twin model in step S5 to form a feature library containing time series data, the system can initiate trend analysis of these features. For example, for the body color features of aquatic organisms, images can be collected periodically, color histograms or average color values ​​of specific regions can be extracted, and these can be stored as part of the individual identification features.

[0090] In step S6, the system can use a moving average algorithm or a linear regression model to analyze the average body surface color over a past period (e.g., the past 7 days or 30 days), calculate its rate of change or trend slope, and thus obtain the trend results of the body surface color characteristics.

[0091] Subsequently, in step S7, the system compares the calculated trend of the body surface color change with a preset range. This preset range can be set based on the aquatic organism's historical health data and the experience of technicians; for example, it can be set as an average body surface color change rate within ±X% per day. If the calculated trend exceeds this preset normal fluctuation range, for example, if the body surface color rapidly lightens or darkens within a short period of time, it indicates an abnormality.

[0092] Finally, in step S8, when the trend of changes in body surface color deviates from the preset range, the system can immediately trigger a health warning. This warning can manifest as sending a text message notification to the aquaculture manager's mobile device, displaying a red alert on the monitoring interface, or automatically adjusting aquaculture environment parameters (e.g., increasing dissolved oxygen or initiating water purification).

[0093] In this way, aquaculture workers can promptly learn about potential health problems of specific aquatic organisms and take swift intervention measures to prevent the spread or worsening of diseases.

[0094] Furthermore, step S1 includes:

[0095] S11: Acquire image data of aquatic organisms from multiple different angles;

[0096] S12: Obtain the outline of aquatic organisms from multiple image data and determine whether the outlines are occluded;

[0097] S13: When the outline is occluded, the occluded outline is compensated based on the unoccluded outline of the aquatic organism and the known morphological features to obtain the complete outline of the aquatic organism.

[0098] S14: Extract macroscopic morphological parameters of aquatic organisms based on the complete outline.

[0099] One approach is to deploy multiple fixed-position camera devices to acquire image data of aquatic organisms from multiple different angles. This multi-angle acquisition strategy significantly increases the possibility of obtaining complete or at least partially unobstructed outlines, providing a rich data source for subsequent processing.

[0100] Determining whether a contour is occluded involves analyzing the integrity, smoothness, and presence of abnormal depressions, breaks, or other features of the extracted aquatic organism contour using existing image processing algorithms. This can be done by calculating the perimeter-to-area ratio of the contour and checking the number of convex hull defects. If the number of convex hull defects exceeds a preset threshold, or if certain parts of the contour exhibit unnatural straight lines or sharp angles, then occlusion can be determined.

[0101] For occluded contours, the system will activate a contour compensation mechanism. Specifically, it can first identify the unoccluded continuous parts of the contour; then, it can use a pre-established statistical shape model of the aquatic organism of this type as a known morphological feature to compensate for the occluded contour.

[0102] The statistical shape model (SSM) was constructed by principal component analysis (PCA) on a large number of unobstructed images of the same aquatic organism, and it contains the main patterns of morphological changes in this species. The statistical shape model has been widely used in medical image analysis by Tim Cootes et al., and the present inventors have innovatively applied this model to the field of aquatic organism morphological analysis, solving the problem of inaccurate morphological parameter measurement caused by occlusion in aquatic organism image recognition.

[0103] During the compensation process, the unoccluded outline can be fitted to a statistical shape model (SSM). The SSM is then used to predict the optimal shape of the occluded portion, thus generating the complete outline of the aquatic organism. For example, for a partially occluded fish with a clear head and tail outline but obscured midsection by aquatic plants, the system can reconstruct the occluded belly and back outlines using the SSM based on the clear head and tail outlines and the typical body shape curve of the fish species. Once the complete outline of the aquatic organism is obtained, its macroscopic morphological parameters can be accurately calculated. For instance, the body surface area can be obtained by calculating the number of pixels in the complete outline, the body length by measuring the length of the longest axis of the outline, and the body width by measuring the maximum width perpendicular to the longest axis. These parameters are then used for subsequent growth stage determination and individual identification.

[0104] This method ensures that the digital twin model can continuously and accurately track the same physical individual, thereby enabling refined and continuous monitoring of the health status of aquatic organisms.

[0105] Furthermore, step S2 includes:

[0106] S21: Obtain the previous macroscopic morphological parameters of aquatic organisms before a preset time period;

[0107] S22: Calculate the individual growth rate of aquatic organisms based on the current macroscopic morphological parameters;

[0108] S23: Determine the growth stage information of aquatic organisms based on their individual growth rate.

[0109] Among them, the previous macroscopic morphological parameters of aquatic organisms before the preset time period refer to the morphological data such as body length, weight, and body width of aquatic organisms recorded at a specific time point or time period before the current time point.

[0110] In some preferred embodiments, this application is implemented as follows:

[0111] In the digital twin model of aquatic organisms, a historical macroscopic morphological parameter database is maintained for each identified physical individual. When it is necessary to determine the growth stage information of an aquatic organism, the system first retrieves the body length and weight data recorded for the aquatic organism 30 days ago from this database as the previous macroscopic morphological parameters for a preset time period.

[0112] Simultaneously, through real-time image acquisition and processing, the system obtains the current body length and weight data of the aquatic organism, serving as its current macroscopic morphological parameters. Next, the system calculates the individual growth rate of the aquatic organism based on these current and previous macroscopic morphological parameters. For example, the body length growth rate can be calculated as (current body length - previous body length) / 30 days, and the weight growth rate can be calculated as (current weight - previous weight) / 30 days. Both body length and weight growth rates can be considered together, or one can be selected as the primary indicator.

[0113] Finally, the system compares the calculated individual growth rate with preset growth stage standards. For example, a growth rate threshold table can be preset, defining the growth stages corresponding to different growth rate ranges (e.g., juvenile stage: body length growth rate greater than X mm / day; subadult stage: body length growth rate between Y and Z mm / day; adult stage: body length growth rate less than W mm / day). The system maps the calculated individual growth rate to this threshold table, thereby determining the current growth stage of the aquatic organism.

[0114] In this way, even if two fish are the same length, the system can determine that they are in different growth stages if their growth rates are different. For example, one may be at the end of the rapid growth juvenile stage, while the other may be in the slow growth adult stage.

[0115] Furthermore, step S23 includes:

[0116] S231: Obtain species information of aquatic organisms and obtain standard growth rate curves of aquatic organisms based on the species information;

[0117] S232: Map the individual growth rate onto the standard growth rate curve to obtain the point of overlap between the individual growth rate and the standard growth rate curve.

[0118] S233: Determine growth stage information based on the overlapping points.

[0119] The standard growth rate curve refers to the trend curve of the growth rate of a specific aquatic organism under ideal growth conditions, which changes with time or body size. Virtual experiments can be conducted in advance to obtain the standard growth rate curve.

[0120] The overlapping point refers to the position point on the standard growth rate curve corresponding to the individual's growth rate. Specifically, it can be determined by calculating the minimum distance between the individual's growth rate and each point on the standard growth rate curve. If the minimum distance is within the preset overlapping threshold range, the overlapping point can be determined.

[0121] When it is necessary to determine the growth stage information of a farmed sturgeon, the system first obtains the species information of the sturgeon, such as "Siberian sturgeon," through a preset database or user input. Based on this species information, the system retrieves the standard growth rate curve of the Siberian sturgeon from the pre-stored database. This curve can be a function curve fitted based on a large amount of historical growth data of Siberian sturgeon, with the horizontal axis representing time or body length and the vertical axis representing growth rate.

[0122] The system then maps the previously calculated individual growth rate of the sturgeon, such as the currently measured daily weight gain, onto the standard growth rate curve of the Siberian sturgeon. This mapping process is achieved by finding the point on the curve that is closest to the current individual growth rate value; this point is the point where the individual growth rate coincides with the standard growth rate curve.

[0123] Finally, the system determines the sturgeon's growth stage based on the specific location of the coincidence point on the standard growth rate curve. For example, if the coincidence point is located in the initial steep upward segment of the standard growth rate curve, the sturgeon can be identified as being in the "rapid growth period of juvenile fish"; if the coincidence point is located in the gentle upward segment in the middle of the curve, it can be identified as being in the "stable growth period of sub-adult fish"; and if the coincidence point S232 is located in the plateau period at the end of the curve, it can be identified as being in the "mature adult stage".

[0124] In this way, even if two Siberian sturgeons have slightly different absolute growth rates, as long as their relative positions on their respective standard curves are similar, they can be judged to be in the same growth stage, thus achieving accuracy in the judgment.

[0125] Furthermore, step S3 includes:

[0126] S31: Based on the species information of aquatic organisms, pre-set the standard biological characteristic set corresponding to different growth stages of that species of aquatic organism;

[0127] S32: Select the set of biological characteristics that correspond to the individual's growth stage from the standard set of biological characteristics;

[0128] S33: Identify the biometric features in the biometric feature set as the priority biometric features for identification.

[0129] The standard biological feature set refers to a complete database of biological features for individual identification that are pre-summarized and stored for different growth stages of a specific species of aquatic organism. It includes descriptions of the unique features of different species of aquatic organisms at various growth stages such as juvenile, subadult, and adult stages, such as body surface spots, patterns, bony plate morphology, fin shape, and body color.

[0130] A biofeature set refers to a subset of biofeatures that are highly correlated with a specific growth stage of an aquatic organism, selected from a standard biofeature set.

[0131] In some preferred embodiments, this application is implemented as follows:

[0132] Suppose the aquatic organism to be identified is a sturgeon, and it has been determined through previous steps that it is currently in the "juvenile stage".

[0133] First, the system pre-defines a set of standard biological characteristics for sturgeon at different growth stages based on the sturgeon species information. For example, for juvenile sturgeon, the standard biological characteristic set might include the distribution of spots on the body surface, the depth of body color, the shape and outline of fins, and the initial number and arrangement pattern of dorsal bony plates; for adult sturgeon, the focus might be more on the complete morphology, number, and arrangement pattern of the bony plates, as well as the fine structure of the body surface markings. These characteristics and their typical manifestations at different stages are stored in a queryable database.

[0134] Subsequently, the system will select a set of biological features corresponding to the current "juvenile stage" of the sturgeon from this preset set of standard biological features. For example, if the sturgeon is currently in the juvenile stage, the system will filter out all features related to the juvenile stage from the set of standard biological features to form a set of biological features, which may include the distribution of spots on the body surface, the depth of body color, and the outline of the fins.

[0135] Finally, the system will identify all the biomarkers in this set as the priority biomarkers for identifying the sturgeon. This means that in subsequent individual identification processes, the system will primarily utilize these identified priority biomarkers to extract individual identification features from image data and perform matching, thereby improving the accuracy and efficiency of identification. For example, when identifying juvenile sturgeon, the system will prioritize analyzing its body surface spots and fin shapes, rather than searching for features on its underdeveloped bony plates. This avoids invalid feature extraction and matching, ensuring the effectiveness of identification.

[0136] Furthermore, step S4 includes:

[0137] S41: Based on biometric priority identification features, extract individual identification features from image data and map the individual identification features and historical individual identification features into the feature measurement space;

[0138] S42: Within the feature metric space, calculate the distance between the mapped individual identification features and the historical individual identification features;

[0139] S43: Based on the distance, determine the matching results between individual identification features and historical individual identification features to identify the physical individual of aquatic organisms.

[0140] The feature metric space is a designed mathematical space where similar individuals are separated by smaller distances of their identifying features, while dissimilar individuals are separated by larger distances, thus enabling the differentiation of individuals. Linear Discriminant Analysis (LDA) can be used to project the original high-dimensional features into a low-dimensional space, allowing feature similarity to be directly reflected by distance. The aim is to optimize feature representation, reduce feature dimensionality, and enhance the discriminative power of features, thereby improving matching efficiency and accuracy.

[0141] Here, distance refers to a numerical value that measures the degree of similarity or difference between two feature vectors within the feature metric space. Specifically, cosine similarity can be used to measure the similarity between individual identification features and historical individual identification features, providing a basis for subsequent matching judgments.

[0142] This application innovatively applies the mature Linear Discriminant Analysis (LDA) algorithm to the field of aquatic organism identification, providing a targeted solution for the continuous and accurate identification of aquatic organisms. By reflecting the individual's identity through distance changes in the feature measurement space, it solves the problems of computational efficiency and misjudgment that exist when directly performing feature matching, ensuring the continuous identification of aquatic organisms and providing a data foundation for subsequent health monitoring and management.

[0143] Furthermore, step S42 includes;

[0144] S421: Based on biometric priority identification features, set weight coefficients for different feature dimensions in the feature measurement space;

[0145] S422: Based on the weighting coefficients, calculate the distance between the mapped individual identification features and the historical individual identification features.

[0146] This application's solution effectively improves the accuracy of aquatic organism identification by introducing a weighted distance calculation mechanism. Specifically, before calculating the distance between the mapped individual identification features and historical individual identification features, weight coefficients are first assigned to different feature dimensions within the feature measurement space based on the biologically preferred identification features. This means that the system assigns higher weights to the corresponding dimensions in the feature measurement space based on pre-determined biological features that have higher discriminative power for individual identification. Conversely, if a feature is prone to change or has small differences between individuals, its corresponding dimension weight will be relatively low.

[0147] In some preferred embodiments, this application is implemented as follows. For example, for a specific aquatic organism, such as an adult sturgeon, its biologically preferential identification characteristics may include the arrangement pattern of its dorsal bony plates, the density distribution of lateral line spots, and the specific shape of its caudal fin. In the feature metric space, these features correspond to different feature dimensions. The system can be pre-set that the bony plate arrangement pattern and caudal fin shape, due to their high stability and distinguishability among adult sturgeon individuals, can be assigned higher weight coefficients to their corresponding feature dimensions; while the lateral line spot density distribution may vary slightly due to environmental factors or minor damage, and its corresponding feature dimension can be assigned relatively lower weight coefficients. For other feature dimensions with lower distinguishability or that are easily variable, such as body color brightness, even lower weight coefficients can be assigned.

[0148] When calculating the distance between the mapped individual identification features and the historical individual identification features, a weighted Euclidean distance can be used. Assume the current individual identification feature vector is X = (x1, x2, ..., x...). n The historical individual identification feature vector is Y = (y1, y2, ..., y3). n The corresponding weight coefficient vector is W = (w1, w2, ..., w n ), where n is the number of individual identification features.

[0149] The weighted distance D(X,Y) can be calculated using the formula... Let's calculate. Where i = 1, 2, ..., n; x i Let y be the value of sample X in the i-th feature dimension; i Let w be the value of sample Y in the i-th feature dimension; i The weight of the i-th feature dimension reflects the importance of that dimension in the similarity measurement.

[0150] Thus, during the calculation process, the difference in each dimension (x) i -y i ) 2 It will be multiplied by its corresponding weight w i This allows even small differences in important feature dimensions to have a significant impact on the final distance, while larger differences in unimportant feature dimensions have a smaller impact, thus ensuring that the matching results more accurately reflect the similarity of biometric priority features.

[0151] Furthermore, step S5 includes:

[0152] S51: Acquisition time for individual identification features;

[0153] S52: Correlate individual identification features with the collection time;

[0154] S53: Store the associated individual identification features in chronological order;

[0155] S54: Append the new individual identification features to the end of the features stored in time sequence to update the feature library of physical individuals.

[0156] Associating individual identification features with collection time means establishing a logical correspondence between each extracted individual identification feature and the specific timestamp when the feature was collected. This can be achieved by binding the timestamp as metadata of the feature data. The purpose is to clarify the time point when each feature was generated, providing a foundation for subsequent time series analysis.

[0157] Storing associated individual identification features in chronological order means arranging and saving individual identification features that have been associated with the collection time according to the order in which they were collected.

[0158] Adding a new individual identification feature to the end of a time-series stored feature means adding a newly acquired individual identification feature associated with a timestamp at the end of an existing sequence of individual identification features arranged in chronological order.

[0159] In some preferred embodiments, specifically, after the system successfully extracts the individual identification features of aquatic organisms through image recognition, it immediately obtains the system timestamp at the time the feature was collected. For example, the shooting time can be read from the metadata of the image file.

[0160] Subsequently, this timestamp is bound to the extracted individual identification feature data; for example, they can be encapsulated into a data object. Then, the system stores this timestamp-bound individual identification feature data in the digital twin model feature library corresponding to the physical individual, according to the chronological order of its collection time. This can be specifically achieved by maintaining a feature table for each physical individual in the database, containing feature data fields and a timestamp field, with the timestamp used as the primary index or sort key.

[0161] When a new individual identification feature is identified and time-linked, the system performs an append operation, inserting the new feature data at the end of the feature table. This ensures that all feature data is arranged in chronological order and that historical data is not overwritten or deleted. For example, if the database already stores feature data for the fish on different dates, the new feature data will be added to the end of the existing data, forming a complete, time-incrementing feature sequence, thereby continuously updating the feature database for that physical individual.

[0162] Please refer to Figure 2 , Figure 3A digital twin image recognition system for aquatic organisms, characterized in that, for implementing any of the above methods, the system comprises:

[0163] Acquisition module 201: Acquires image data of aquatic organisms and extracts macroscopic morphological parameters of aquatic organisms based on the image data;

[0164] Judgment Module 202: Determine the growth stage information of aquatic organisms based on macroscopic morphological parameters;

[0165] Module 203: Based on growth stage information, determine the priority identification biological characteristics of aquatic organisms;

[0166] Identification module 204: Based on priority identification biological features, extract individual identification features of aquatic organisms from image data, and match the individual identification features with historical individual identification features stored in the digital twin model to identify the physical individual of the aquatic organism;

[0167] Update module 205: Update individual identification features to the digital twin model to update the feature library of the physical individual.

[0168] The acquisition module 201 refers to the functional unit used to receive and process aquatic organism image data and extract macroscopic morphological parameters from it. It can be implemented using image acquisition equipment, image processing software or hardware circuits.

[0169] The judgment module 202 refers to the functional unit used to analyze and determine the current growth stage information of aquatic organisms based on macroscopic morphological parameters.

[0170] The determination module 203 refers to the functional unit used to select and specify the priority identification biological characteristics of aquatic organisms applicable to the current stage based on the determined growth stage information. It can be implemented using a feature library management system, feature selection algorithm or expert system.

[0171] The matching module 204 refers to a functional unit used to extract individual identification features of aquatic organisms from image data based on priority identification biological features, and compare them with historical individual identification features stored in the digital twin model to confirm the physical identity of aquatic organisms. It can be implemented using pattern recognition algorithms, deep learning networks or feature vector distance calculators.

[0172] The update module 204 refers to the functional unit used to integrate and store newly extracted individual identification features into the digital twin model in order to continuously improve and update the physical individual feature library. It can be implemented using a database management system, a data synchronization mechanism, or a feature fusion algorithm.

[0173] The solution in this application adopts a modular design, in which each step of the image recognition method for digital twins of aquatic organisms is implemented by a corresponding functional module, thereby providing a system corresponding to the method.

[0174] Specifically, the system first receives image data of aquatic organisms through the acquisition module 201, and performs preliminary processing on this data to extract macroscopic morphological parameters of the aquatic organisms, laying the data foundation for subsequent analysis. Then, the judgment module 202 receives these macroscopic morphological parameters and automatically determines the current growth stage information of the aquatic organisms based on preset logic or models. Based on this growth stage information, the determination module 203 intelligently selects and specifies the most suitable priority biological features for identification at the current stage, ensuring the accuracy and robustness of the identification.

[0175] Next, the matching module 204 uses these priority biological features to accurately extract the individual identification features of aquatic organisms from the original image data and compares them with the historical individual identification features stored in the digital twin model, thereby realizing the identification of physical individuals of aquatic organisms.

[0176] Finally, the update module 205 is responsible for updating the newly identified individual features to the corresponding digital twin model in a timely manner, continuously improving the feature library of physical individuals.

[0177] Through this systematic implementation, this application not only concretizes abstract methods and steps into operable functional modules, making the entire identification process more stable and controllable, but also ensures the continuity and accuracy of aquatic organism identification through close collaboration and data flow between modules. This system-level protection effectively avoids the circumvention problems that may exist if relying solely on methodological procedures, enabling continuous long-term tracking, health monitoring, and trend analysis of aquatic organisms, thus providing solid technical support for the refined management of aquaculture.

[0178] In this document, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, without necessarily requiring or implying any such actual relationship or order between these entities or operations.

[0179] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for image recognition of digital twins of aquatic organisms, characterized in that, The method includes the following steps: S1: Acquire image data of aquatic organisms, and extract macroscopic morphological parameters of the aquatic organisms based on the image data; S2: Determine the growth stage information of the aquatic organism based on the macroscopic morphological parameters; S3: Based on the growth stage information, determine the priority identification biological characteristics of the aquatic organism; Step S3 includes: S31: Obtain information on the types of aquatic organisms, and based on the information on the types of aquatic organisms, pre-set a set of standard biological characteristics corresponding to the types of aquatic organisms at different growth stages. S32: Select the set of biometric features corresponding to the growth stage information from the standard set of biometric features; S33: The biometric features in the set of biometric features are identified as the priority biometric features; S4: Based on the prioritized biological characteristics, extract the individual identification features of the aquatic organism from the image data, and match the individual identification features with the historical individual identification features stored in the digital twin model to identify the physical individual of the aquatic organism; S5: Update the individual identification features to the digital twin model to update the feature library of the physical individual.

2. The image recognition method for digital twins of aquatic organisms according to claim 1, characterized in that, Step S5 and the following steps include: S6: Based on the updated feature library, analyze the changing trend of the individual identification features over time to obtain the changing trend results; S7: Compare the changing trend results with the preset range; S8: When the change trend result deviates from the preset range, a health warning is triggered.

3. The image recognition method for digital twins of aquatic organisms according to claim 1, characterized in that, Step S1 includes: S11: Acquire the image data of the aquatic organism from multiple different angles; S12: Obtain the outline of the aquatic organism from multiple image data, and determine whether the outline is occluded; S13: When the outline is occluded, the occluded outline is compensated based on the unoccluded outline of the aquatic organism and known morphological features to obtain the complete outline of the aquatic organism. S14: Extract the macroscopic morphological parameters of the aquatic organism based on the complete outline.

4. The image recognition method for digital twins of aquatic organisms according to claim 1, characterized in that, Step S2 includes: S21: Obtain the previous macroscopic morphological parameters of the aquatic organisms before a preset time period; S22: Calculate the individual growth rate of the aquatic organism based on the previously obtained macroscopic morphological parameters and the macroscopic morphological parameters; S23: Determine the growth stage information of the aquatic organism based on the individual growth rate.

5. The image recognition method for digital twins of aquatic organisms according to claim 4, characterized in that, Step S23 includes: S231: Obtain the species information of the aquatic organism, and obtain the standard growth rate curve of the aquatic organism based on the species information; S232: Map the individual growth rate onto the standard growth rate curve to obtain the point of overlap between the individual growth rate and the standard growth rate curve; S233: Determine the growth stage information based on the overlapping points.

6. The image recognition method for digital twins of aquatic organisms according to claim 1, characterized in that, Step S4 includes: S41: Based on the priority biometric features, extract individual identification features from the image data, and map the individual identification features and the historical individual identification features into the feature metric space; S42: Within the feature metric space, calculate the distance between the mapped individual identification feature and the historical individual identification feature; S43: Based on the distance, determine the matching result between the individual identification feature and the historical individual identification feature to identify the physical individual of the aquatic organism.

7. The image recognition method for digital twins of aquatic organisms according to claim 6, characterized in that, Step S42 includes; S421: Based on the biometric priority identification features, set weight coefficients for different feature dimensions in the feature measurement space; S422: Based on the weighting coefficients, calculate the distance between the mapped individual identification features and the historical individual identification features.

8. The image recognition method for digital twins of aquatic organisms according to claim 1, characterized in that, Step S5 includes: S51: Obtain the acquisition time of the individual identification features; S52: Associate the individual identification features with the collection time; S53: Store the associated individual identification features in chronological order; S54: Append the individual identification features to the end of the features stored in time sequence to update the feature library of the physical individual.

9. A digital twin image recognition system for aquatic organisms, characterized in that, The system for implementing the method according to any one of claims 1-8 comprises: Acquisition module: Acquires image data of aquatic organisms and extracts macroscopic morphological parameters of the aquatic organisms based on the image data; Judgment module: Based on the macroscopic morphological parameters, determine the growth stage information of the aquatic organism; Determination module: Based on the growth stage information, determine the priority identification biological characteristics of the aquatic organism; The determining module is further configured to: acquire species information of aquatic organisms; pre-set a set of standard biological characteristics corresponding to different growth stages of aquatic organisms based on the species information of aquatic organisms; select a set of biological characteristics corresponding to the growth stage information from the set of standard biological characteristics; and determine the biological characteristics in the set of biological characteristics as the priority identification biological characteristics. Identification module: Based on the priority biological features to be identified, extract the individual identification features of the aquatic organism from the image data, and match the individual identification features with the historical individual identification features stored in the digital twin model to identify the physical individual of the aquatic organism; Update module: Updates the individual identification features to the digital twin model to update the feature library of the physical individual.

Images