Method and system for measuring high-throughput morphology of fish

Abstract

The invention belongs to the technical field of high-throughput morphology determination, and discloses a fish high-throughput morphology determination method, which has the advantages of high detection speed and high accuracy: a deep neural network is used for analyzing an acquired image, compared with the traditional manual measurement, the detection speed is higher, the detection is more accurate, the detection speed reaches 1 second per fish, and the detection accuracy is higher. Compared with manual detection, the detection is improved by more than 10 times, and the accuracy is up to Compared with the prior art, testing steps and testing tools are reduced, the projection geometry principle is utilized, the fish body is placed under the image collector, the direction vector of the fish body is calculated in real time through a program, the accurate placing angle is automatically given, and the vertical collection image of the fish body can be obtained after operation is carried out according to program prompts. And processing the image through graying and Gaussian processing to obtain the vertical projection of the fish body. And the difficulty and error of the body curved surface test are reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of high-throughput morphological measurement, and particularly relates to a method and system for high-throughput morphological measurement of fish. Background Technique

[0002] Currently, in fish breeding operations, the measurement of measurable traits of broodstock relies on traditional manual measurement of morphology. By measuring traits such as the weight, total length, body length, body height, eye diameter, etc. of broodstock, the traits of broodstock can be evaluated. In measurement, the stability of data and the speed of measurement are important prerequisites for ensuring the quality and efficiency of measurement data. During the process of fish breeding operations, the number of broodstock is large and the subjectivity of manual measurement is high. There is still a lack of automated and standardized measurement technical solutions.

[0003] Performing high-throughput measurement of morphological indices such as the total length, body length, body height, head length, snout length, caudal peduncle length, caudal peduncle width, tail length, weight, etc. of broodstock based on machine vision technology is a research direction that breeding operators are keen on. However, various unqualified data will be obtained due to environmental problems during the image acquisition process. For example, the body shapes of different species of fish vary greatly, and there is no overall solution yet; during the process of fish measurement, factors such as residual water stains, light, body color, etc. all affect the collected data, generating a large number of redundant invalid images. Therefore, there is an urgent need for a method for high-throughput morphological measurement of fish to eliminate a large number of unqualified generated images and automatically obtain morphological data to ensure the accuracy and reliability of the output results of measurement data.

[0004] Through the above analysis, the problems and defects existing in the prior art are as follows:

[0005] Performing high-throughput measurement of morphological indices such as the total length, body length, body height, head length, snout length, caudal peduncle length, caudal peduncle width, tail length, weight, etc. of broodstock based on machine vision technology is a research direction that breeding operators are keen on. However, various unqualified data will be obtained due to environmental problems during the image acquisition process. Summary of the Invention

[0006] Aiming at the problems existing in the prior art, the present invention provides a method for high-throughput morphological measurement of fish.

[0007] The present invention is implemented as follows. A method for high-throughput morphological measurement of fish includes:

[0008] (1) Collect fish body images through an image shooting auxiliary component and introduce a standard size reference object;

[0009] (2) Use image recognition software to process the collected fish body images, including data preprocessing, feature extraction, target detection, and morphological measurement;

[0010] (3) Construct a morphological measurement model based on a deep learning algorithm to calculate and analyze the morphological data of the fish body;

[0011] (4) Store, calculate, and output the measurement data through the backend integration program;

[0012] (5) Display the morphological measurement results on the WEB side and support real-time image acquisition and adjustment prompts.

[0013] Furthermore, the image shooting auxiliary component includes:

[0014] (1) An integrally formed fish body test background board, which includes a special background color and a standard reference object;

[0015] (2) A weight measurement induction single board computer for measuring the weight of the fish body;

[0016] (3) A stable and adjustable support rod for fixing and height adjustment of the camera;

[0017] (4) A high-definition camera with adjustable parameters, including exposure time and focal length;

[0018] (5) A bottom plate structure for fixing the support rod and other components;

[0019] (6) A supplementary lighting device and connectors.

[0020] Furthermore, the image recognition software runs on a computing platform that meets the following hardware configurations:

[0021] (1) Windows 10 operating system;

[0022] (2) An Intel i5-13500H or higher CPU;

[0023] (3) An independent graphics card GPU;

[0024] (4) 8GB or more of memory;

[0025] (5) An SSD with 512GB or more of storage space.

[0026] Furthermore, the image recognition software includes:

[0027] (1) A target detection algorithm based on deep learning;

[0028] (2) A dataset for training the target detection model;

[0029] (3) A real-time image detection and processing module;

[0030] (4) A backend integration program that implements data processing and user interaction based on the Python Flask framework.

[0031] Furthermore, the target detection algorithm adopts a fully convolutional neural network architecture and includes the following calculation steps:

[0032] (1) Divide the input image into grids, and each grid is used to detect the target area;

[0033] (2) Calculate multiple bounding boxes and their confidence scores;

[0034] (3) Predict the conditional probability of the target category;

[0035] (4) Calculate the final target recognition result based on the confidence score and the category probability.

[0036] Furthermore, the back-end integration program includes a WEB service for processing user test requests, and the specific process of the WEB service includes:

[0037] (1) The user submits a fish body morphology measurement request on the WEB interface;

[0038] (2) The back-end calls the camera to collect real-time images and returns them to the WEB interface;

[0039] (3) The back-end calls the deep learning algorithm for target detection to identify the fish body and the standard reference object;

[0040] (4) If the position of the fish body deviates, the system gives an adjustment prompt;

[0041] (5) When the fish body is placed correctly, the back-end automatically calculates the morphological measurement results and displays them on the WEB interface.

[0042] Furthermore, the measurement process on the WEB interface includes the following operation steps:

[0043] (1) The user places the fish body on the fish body test background board;

[0044] (2) The system collects real-time images and determines whether the position of the fish body meets the measurement standard;

[0045] (3) When the position of the fish body is correct, the system automatically calculates morphological data such as body length, body length, head length, snout length, caudal peduncle length, caudal peduncle width, tail length, and body weight;

[0046] (4) The measurement results are stored in the database and support data export and remote access.

[0047] Furthermore, a high-throughput fish morphology measurement system includes:

[0048] (1) An image acquisition module, which is used to capture fish body images through a high-definition camera and introduce a standard-size reference object for scale calibration;

[0049] (2) A data processing module, including a deep learning target detection algorithm, a feature extraction algorithm, and a morphology calculation model, analyzes the collected images and calculates fish body morphology parameters;

[0050] (3) The back-end integration module, based on the Python Flask framework, is responsible for data storage, morphology calculation, and user interaction;

[0051] (4) The WEB interaction module is used to receive user measurement requests, return measurement results, and provide real-time adjustment prompts;

[0052] (5) The hardware support module includes a fish body test background board, an adjustable high-definition camera, a weight measurement induction single-board computer, a stable support rod, and a supplementary lighting device.

[0053] Further, the image acquisition module includes:

[0054] (1) An integrally formed fish body test background board, including a specific color background and a standard reference object;

[0055] (2) An adjustable high-definition camera, supporting the adjustment of exposure time and focal length;

[0056] (3) A stable support rod with adjustable height for fixing the camera;

[0057] (4) A weight measurement induction single-board computer for measuring the weight of the fish body;

[0058] (5) A supplementary lighting device to improve the measurement accuracy in low-light environments.

[0059] Further, the data processing module includes:

[0060] (1) A target detection algorithm based on deep learning, adopting a fully convolutional neural network architecture;

[0061] (2) A data training set for fish body morphology determination;

[0062] (3) A morphology calculation model for extracting features from fish body images and calculating morphological parameters such as body length, body depth, head length, snout length, caudal peduncle length, caudal peduncle width, tail length, and weight;

[0063] (4) A data storage and management module, supporting database storage, measurement record retrieval, and data export;

[0064] (5) A back-end interaction program, integrated into the WEB server, supporting remote access to measurement data by users.

[0065] Combining the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solution to be protected by the present invention are:

[0066] First, the present invention uses a deep neural network to automatically analyze fish body images. Compared with traditional manual measurement methods, the detection speed is faster and the accuracy is higher. Through one scan, eight indicators such as body length, body height, body width, and body weight can be obtained simultaneously. While it takes at least 2 minutes to measure the same data manually, the present invention only takes 1 second, and the efficiency is increased by more than 120 times. Thanks to the optimization of the projection geometry principle, this system only needs to simply place the fish body to achieve accurate measurement, avoiding multiple test steps and the use of complex tools in manual operation. This enables operators without professional backgrounds to quickly get started, significantly reducing the learning cost and the usage threshold.

[0067] Traditional fish morphology measurement methods rely heavily on manual experience. They not only require high professional qualities of operators, but also have great measurement difficulties and strong resistance from inspectors due to the slippery fish body and the need for repeated operations. Especially in large-scale sample measurements, both labor costs and errors will increase exponentially. The present invention adopts an integrated detection scheme combining single-lens scanning and a gravity sensor. Only one operator is required to complete the detection, without additional training or complex skills, significantly reducing the requirements for personnel's education level and physical strength, effectively reducing data deviation caused by human errors, reducing detection costs, and improving data stability.

[0068] The present invention can not only measure 9 morphological data such as total fish length, body length, body height, head length, snout length, caudal peduncle length, caudal peduncle height, tail length, and body weight at one time, but also can meet more biological measurement requirements through algorithm optimization and hardware expansion, such as detecting the scale size, fin length, etc. The system is applicable to various water environments and can optimize the lighting conditions through lighting equipment, enabling it to still work stably in complex environments such as low light and strong light. In addition, this detection scheme can adjust the measurement parameters according to the morphological characteristics of different fish species, making it have a wider applicability and meeting the needs of multiple industries such as scientific research, fishery management, and aquaculture evaluation.

[0069] The present invention adopts data intelligent storage and management technology. The detection results can be directly stored in the database and support functions such as audit records and data traceability to ensure the integrity and traceability of the data. At the same time, the system supports multiple data export formats such as CSV and EXCEL, which is convenient for scientific researchers to quickly organize and analyze, meeting the needs of big data calculation and statistics. Compared with the traditional manual recording method, this system realizes automatic data storage and real-time update, avoids human input errors, improves data utilization efficiency, and facilitates team sharing and remote access, making fish morphology research more intelligent and efficient.

[0070] Second, the technical solution of the present invention has significant value in industrial application, especially in improving scientific research efficiency, reducing measurement costs, and optimizing data quality. The traditional manual measurement method is inefficient due to the slippery fish body and complex operations, taking several minutes for each measurement. When conducting large-scale measurements, researchers face a huge workload and data errors. The present invention adopts high-throughput automated detection technology, which can complete a full measurement in only 1 second, with an efficiency improvement of more than 100 times. This enables researchers to complete more experiments in a shorter time and improve the progress of scientific research. For example, in an experiment of measuring 1,000 fish, the traditional method takes 33 hours, while the solution of the present invention only takes 17 minutes, greatly shortening the experimental cycle and improving the data collection efficiency. At the same time, the solution simplifies the detection steps to a single placement operation, enabling non-professionals to easily get started, reducing the personnel training cost, and enhancing the popularization of the technology.

[0071] The technical solution of the present invention reduces the dependence on manual measurement during the experiment through automated detection means, thus significantly reducing the personnel cost and the burden of repetitive labor. In the traditional method, at least two staff members are required to complete the measurement per hour, while the efficient detection system of the present invention only requires one operator to complete the same task, reducing the labor cost by more than 50%. Especially in large-scale experiments or long-term data collection projects, the saved labor cost can reach hundreds of thousands of yuan. At the same time, the system reduces the hygiene and safety risks brought by experimental personnel's contact with fish bodies through non-contact measurement methods, avoiding occupational health problems caused by operation errors or long-term contact with biological samples. In addition, researchers no longer need to measure repeatedly, reducing data errors caused by fatigue, improving the reliability of the data, and reducing the psychological pressure and resistance of the working environment on personnel, making scientific research experiments more efficient and user-friendly.

[0072] The present invention has achieved a breakthrough innovation in the field of fish morphology determination, filling the technical gap in this field at home and abroad. The traditional measurement method relies on manual operation. Although it can obtain a certain degree of accuracy, it is limited by subjective errors, environmental impacts, and data consistency issues, and it is difficult to meet the needs of modern scientific research for large-scale data collection. The present invention first introduces deep learning technology into fish morphology determination, combines computer vision, data analysis, and intelligent measurement technology, and realizes the high-precision and high-throughput automatic determination of fish body morphology data, significantly improving the reliability and scalability of data acquisition. In addition, through automated optical imaging and data calculation, the technical solution realizes the simultaneous determination of multiple morphological parameters such as fish body length, body height, head length, and tail length, and can expand more indicators according to scientific research needs. This technological breakthrough not only has wide application value in the academic research field but also provides advanced data collection means for industries such as aquaculture and environmental monitoring, and provides a new scientific tool for fish ecology research and biodiversity conservation.

[0073] Traditional fish measurement techniques have long relied on manual operations. Although accurate data can be obtained to a certain extent, it is difficult to meet the requirements of modern, large-scale, and precise measurements due to the slippery and water-stained fish bodies, as well as the subjective judgment errors of personnel. Especially for big data research, the inefficiency of traditional methods has become a major bottleneck in the scientific research process. The technical solution of this invention breaks through this limitation through deep learning, computer vision, and sensor fusion technologies, achieving precise, high-speed, and large-scale automated acquisition of fish morphological data. This solution can not only be used in laboratory environments but also be extended to multiple fields such as fishery production, environmental protection, and aquatic biological resource surveys, providing an efficient and standardized data acquisition method, enabling fish morphological measurement to move from the traditional manual experience mode to a data-driven intelligent stage, and providing important support for the technological upgrading of related industries worldwide. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 is a flowchart of the method for high-throughput fish morphological measurement provided by an embodiment of the present invention.

[0075] Figure 2 is a block diagram of the system structure for high-throughput fish morphological measurement provided by an embodiment of the present invention.

[0076] Figure 3 is a structural diagram of the device for high-throughput fish morphological measurement provided by an embodiment of the present invention.

[0077] Figure 4 is a side view of the device for high-throughput fish morphological measurement provided by an embodiment of the present invention.

[0078] Figure 5 is a rear view of the device for high-throughput fish morphological measurement provided by an embodiment of the present invention.

[0079] In the figure: 1, bottom plate; 2, buckle plate; 3, gravity sensor; 4, single-chip microcomputer; 5, main support rod; 6, camera support rod; 7, camera; 8, standard measuring piece. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0080] In order to make the objectives, technical solutions, and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below in conjunction with embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0081] As Figure 1 shown, a method for high-throughput fish morphological measurement provided by an embodiment of the present invention includes the following steps:

[0082] S101, collect fish body images through an image shooting auxiliary component and introduce a standard size reference object;

[0083] S102. Process the collected fish body images using image recognition software, including data preprocessing, feature extraction, target detection, and morphological measurement;

[0084] S103. Build a morphological measurement model based on deep learning algorithms to calculate and analyze the morphological data of the fish body;

[0085] S104. Store, calculate, and output the measurement data through a backend integration program;

[0086] S105. Display the morphological measurement results on the WEB side and support real-time image acquisition and adjustment prompts.

[0087] The image shooting auxiliary component provided by the embodiment of the present invention includes:

[0088] (1) An integrally formed fish body test background board, which includes a special background color and a standard reference object;

[0089] (2) A weight measurement induction single board computer for measuring the weight of the fish body;

[0090] (3) A stable and adjustable support rod for fixing and height adjustment of the camera;

[0091] (4) A high-definition camera with adjustable parameters, including exposure time and focal length;

[0092] (5) A bottom plate structure for fixing the support rod and other components;

[0093] (6) A supplementary lighting device and connectors.

[0094] The image recognition software provided by the embodiment of the present invention runs on a computing platform that meets the following hardware configurations:

[0095] (1) Windows 10 operating system;

[0096] (2) An Intel i5-13500H or higher CPU;

[0097] (3) An independent graphics card GPU;

[0098] (4) 8GB or more of memory;

[0099] (5) An SSD with a storage space of 512GB or more.

[0100] The image recognition software provided by the embodiment of the present invention includes:

[0101] (1) A target detection algorithm based on deep learning;

[0102] (2) A dataset for training the target detection model;

[0103] (3) Real-time image detection and processing module;

[0104] (4) Back-end integration program, which implements data processing and user interaction based on the Python Flask framework.

[0105] The object detection algorithm provided by the embodiment of the present invention adopts a fully convolutional neural network architecture and includes the following calculation steps:

[0106] (1) Divide the input image into grids, and each grid is used to detect the target area;

[0107] (2) Calculate multiple bounding boxes and their confidence scores;

[0108] (3) Predict the conditional probability of the target class;

[0109] (4) Calculate the final target recognition result based on the confidence score and the class probability.

[0110] The back-end integration program provided by the embodiment of the present invention includes a WEB service for processing user test requests, and the specific process of the WEB service includes:

[0111] (1) The user submits a fish body morphology measurement request on the WEB interface;

[0112] (2) The back-end calls the camera to collect real-time images and returns them to the WEB interface;

[0113] (3) The back-end calls the deep learning algorithm for object detection to identify the fish body and the standard reference object;

[0114] (4) If the position of the fish body deviates, the system gives an adjustment prompt;

[0115] (5) When the fish body is placed correctly, the back-end automatically calculates the morphological measurement results and displays them on the WEB interface.

[0116] The WEB interface measurement process provided by the embodiment of the present invention includes the following operation steps:

[0117] (1) The user places the fish body on the fish body test background board;

[0118] (2) The system collects real-time images and judges whether the position of the fish body meets the measurement standard;

[0119] (3) When the position of the fish body is correct, the system automatically calculates morphological data such as body length, body length, head length, snout length, caudal peduncle length, caudal peduncle width, tail length and body weight;

[0120] (4) The measurement results are stored in the database and support data export and remote access.

[0121] Such as Figure 2As shown in the figure, a high-throughput fish morphology measurement system provided by an embodiment of the present invention includes:

[0122] (1) An image acquisition module, which is used to capture fish body images through a high-definition camera and introduce a standard-size reference object for scale calibration;

[0123] (2) A data processing module, including a deep learning object detection algorithm, a feature extraction algorithm, and a morphology calculation model, which analyzes the collected images and calculates fish body morphology parameters;

[0124] (3) A back-end integration module, based on the Python Flask framework, responsible for data storage, morphology calculation, and user interaction;

[0125] (4) A WEB interaction module, which is used to receive user measurement requests, return measurement results, and provide real-time adjustment prompts;

[0126] (5) A hardware support module, including a fish body test background board, an adjustable high-definition camera, a weight measurement induction single-board computer, a stable support rod, and a supplementary lighting device.

[0127] The image acquisition module provided by the embodiment of the present invention includes:

[0128] (1) An integrally formed fish body test background board, which includes a specific color background and a standard reference object;

[0129] (2) An adjustable high-definition camera, which supports adjusting the exposure time and focal length;

[0130] (3) An adjustable-height stable support rod, which is used for camera fixation;

[0131] (4) A weight measurement induction single-board computer, which is used to measure the weight of the fish body;

[0132] (5) A supplementary lighting device, which improves the measurement accuracy in low-light environments.

[0133] The data processing module provided by the embodiment of the present invention includes:

[0134] (1) A deep learning-based object detection algorithm, which adopts a fully convolutional neural network architecture;

[0135] (2) A data training set for fish body morphology measurement;

[0136] (3) A morphology calculation model, which extracts features from fish body images and calculates morphology parameters such as body length, body length, head length, snout length, caudal peduncle length, caudal peduncle width, tail length, and body weight;

[0137] (4) A data storage and management module, which supports database storage, measurement record backtracking, and data export;

[0138] (5) The backend interaction program is integrated into the WEB server and supports users to remotely access measurement data.

[0139] The image-assisted shooting component provided by the embodiment of the present invention includes the following sub-components:

[0140] An integrally formed fish body test background board, the test board contains a special background color and standard reference objects, which can greatly improve the efficiency of deep learning;

[0141] A weight measurement induction single board computer for measuring the weight of the fish body;

[0142] A stable and adjustable support rod for adjusting the height of the camera;

[0143] A high-definition camera with adjustable parameters, and the adjustable parameters include exposure time and focal length;

[0144] A bottom plate for fixing the support rod;

[0145] Light supplement equipment and connectors;

[0146] Other connectors.

[0147] The hardware required for the image recognition software provided by the embodiment of the present invention includes:

[0148] A hardware platform installed with Window 10;

[0149] The CPU is required to be above Inter 13I5-13500H;

[0150] The graphics card is required to be an independent graphics card;

[0151] The memory is 8G or above;

[0152] The hard disk is SSD 512 or above.

[0153] The image recognition software provided by the embodiment of the present invention includes:

[0154] A target detection algorithm based on deep learning;

[0155] Model training data;

[0156] Program software such as real-time detection;

[0157] The target detection algorithm based on deep learning adopts a fully convolutional neural network architecture, and its core calculation formulas include convolution operation, pooling operation and non-linear activation function.

[0158] The main steps of the target detection provided by the embodiment of the present invention are as follows:

[0159] Input image grid division: Divide the input image into a grid of SxS (where S represents the number of image pixels), and each grid is responsible for detecting one target;

[0160] Bounding box prediction: Each grid predicts B bounding boxes and confidence scores; The confidence score represents the probability that the target is contained within the bounding box and the accuracy of the bounding box;

[0161] Class prediction: Each grid also predicts the conditional probabilities of C classes;

[0162] Joint prediction: Combine the confidence score and the class conditional probability to calculate the final confidence score for each bounding box;

[0163] The backend integration program uses the Flask WEB framework in Python and integrates a graphic acquisition device, a target detection algorithm for deep learning, and a WEB service for corresponding user test requests.

[0164] The specific process of the WEB service for corresponding user test requests provided by the embodiments of the present invention is as follows;

[0165] The user sends a fish body index test request on the WEB browser;

[0166] The backend integration program calls the high-definition camera and captures an image and returns it to the user's WEB side;

[0167] The backend integration program calls the deep learning algorithm to identify the captured actual image and confirm whether the fish body is recognized and the standard reference piece;

[0168] If the position of the fish body is not accurately placed, the backend integration program will prompt an error message, such as adjusting the fish body angle information;

[0169] If the fish body is accurately placed, the backend integration program will redraw the image and give a green prompt;

[0170] After the backend measurement is completed, the specific test indicators are displayed on the WEB page, body length, body length, head length, fish snout length, caudal peduncle length, caudal peduncle width, tail length, weight index;

[0171] The specific operation steps are as follows:

[0172] Place the fish body to be tested on the fish body test board, and the program will display a real-time image;

[0173] At this time, the program software will give a prompt according to the real-time captured image. If it is displayed in green, it means that the fish body is placed correctly and the test is successful;

[0174] The program software will display the specific fish body indicators.

[0175] Another object of the present invention is to provide a computer device, which includes a memory and a processor. The memory stores a computer program. When the computer program is executed by the processor, the processor is caused to execute the steps of the method for high-throughput fish morphology determination.

[0176] Another object of the present invention is to provide a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor is caused to execute the steps of the method for high-throughput fish morphology determination.

[0177] Another object of the present invention is to provide an information data processing terminal for implementing the high-throughput fish morphology determination system.

[0178] As Figure 3 、 Figure 4 、 Figure 5 As shown, a high-throughput fish morphology determination device provided by an embodiment of the present invention includes: a bottom plate 1, a buckle plate 2, a gravity sensor 3, a single-chip microcomputer 4, a main support rod 5, a camera support rod 6, a camera 7, and a standard measuring part 8.

[0179] A main support rod 5 is fixed above the right side of the bottom plate 1 by screws, and the bottom of the main support rod 5 is sleeved with the buckle plate 2; a gravity sensor 3 is embedded at the bottom of the buckle plate 2; a single-chip microcomputer 4 is fixed at the bottom of the gravity sensor 3 by screws; a camera support rod 6 is fixed above the main support rod 5 by screws; a camera 7 is fixed at the left end of the camera support rod 6 by screws; a standard measuring part 8 is arranged on the left side of the camera support rod 6.

[0180] The high-throughput fish morphology determination device of the present invention is composed of a bottom plate 1, a buckle plate 2, a gravity sensor 3, a single-chip microcomputer 4, a main support rod 5, a camera support rod 6, a camera 7, and a standard measuring part 8, forming a set of devices that can accurately measure fish morphology parameters. The bottom plate 1 provides stable support for the device, and the main support rod 5 is used to fix the measurement components, enabling the camera 7 to capture the fish body morphology under stable conditions. The camera support rod 6 realizes precise adjustment of the camera angle through connection with the main support rod 5, ensuring that the acquired images are clear and have a unified reference scale.

[0181] The buckle plate 2 is used to fix the fish body to be measured, and a gravity sensor 3 is integrated at its bottom. After the fish body is placed, the weight data can be immediately collected. The gravity sensor 3 is connected to the single-chip microcomputer 4 and transmits the collected weight data to the data processing module to realize the real-time weighing function. In addition, the buckle plate 2 ensures the stability of the fish body, reduces measurement errors, and makes the morphological analysis more accurate.

[0182] The camera 7 is kept stable by being fixed on the camera support rod 6 and captures the morphological characteristics of the fish body, such as body length, body width, fin ray length, etc. During the camera shooting process, the standard measurement piece 8 provides a reference scale to ensure high-precision proportional conversion during subsequent image analysis to obtain accurate fish body size data. The introduction of the standard measurement piece 8 reduces measurement errors and improves data consistency.

[0183] The collected weight data and morphological images are integrated and stored by the single-chip microcomputer 4. The single-chip microcomputer 4 can analyze the signals of the gravity sensor 3 and generate complete morphological measurement information of the fish body in combination with the image data. Subsequently, the data can be uploaded to a computer or database for further analysis and comparative research, thereby providing reliable data support for fish growth assessment, population classification, and breeding research.

[0184] This device combines gravity measurement, standard scale calibration, and high-resolution image acquisition technologies, which can effectively improve the efficiency and accuracy of fish morphological measurement and make it applicable to various application scenarios such as scientific research, aquaculture, and ecological monitoring. This technical solution is an integrated fish body measurement solution, which is simple and easy to operate. Just place the fish body to be measured on the instrument, and it can automatically measure indicators such as total length, body length, body height, head length, snout length, caudal peduncle, caudal peduncle, tail length, and weight.

[0185] The components of this technical solution can be divided into two major parts: the image shooting auxiliary component and the image recognition calculation software. Among them, the image shooting auxiliary component is used to shoot high-quality images and introduces a reference object component with a standard size to explain the specific size in reality represented by one pixel of the captured image, reducing errors caused by graphic distortion and deformation; the image recognition software includes deep learning algorithms, model data, backend integration programs, and front-end interfaces to ensure the simplicity and accuracy of user operations.

[0186] The image-assisted shooting component includes the following sub-components:

[0187] 1. An integrally formed fish body test background board. The test board includes a special background color and a standard reference object, which can greatly improve the efficiency of deep learning.

[0188] 2. A weight measurement induction single-board computer for measuring the weight of the fish body.

[0189] 3. A stable and adjustable support rod for adjusting the height of the camera to ensure the stability of the captured image.

[0190] 4. A high-definition camera with adjustable parameters. The adjustable parameters include exposure time, focal length, etc.

[0191] 5. A bottom plate for fixing the support rod.

[0192] 6. Supplementary lighting device and connecting piece.

[0193] 7. Other connecting pieces.

[0194] Final effect diagram of the image-assisted shooting component.

[0195] The hardware required for the image recognition software includes:

[0196] 1. A hardware platform installed with Window 10.

[0197] 2. The CPU is required to be above Inter 13I5-13500H.

[0198] 3. The graphics card is required to be an independent graphics card.

[0199] 4. The memory is 8G or above.

[0200] 5. The hard disk is SSD 512 or above.

[0201] The image recognition software includes:

[0202] 1. Object detection algorithm based on deep learning.

[0203] 2. Model training data.

[0204] 3. Program software such as real-time detection.

[0205] The object detection algorithm based on deep learning adopts a fully convolutional neural network architecture, and its core calculation formulas include convolutional operation, pooling operation and non-linear activation function. The main steps of object detection are as follows:

[0206] 1. Input image grid division: Divide the input image into grids of SxS (S represents the number of image pixels), and each grid is responsible for detecting one object.

[0207] 2. Bounding box prediction: Each grid predicts B bounding boxes and confidence scores. The confidence score represents the probability that the object is contained in the bounding box and the accuracy of the bounding box.

[0208] 3. Class prediction: Each grid also predicts the conditional probability of C classes.

[0209] 4. Joint prediction: Combine the confidence score and the class conditional probability to calculate the final confidence score of each bounding box.

[0210] The backend integration program adopts the Flask WEB framework of Python, integrates the graphic acquisition device, the object detection algorithm based on deep learning, and the WEB service for corresponding user test requests. The specific process is as follows.

[0211] 1. The user sends a fish body index test request on the WEB browser.

[0212] 2. The backend integration program calls the high-definition camera and captures images for return to the user's WEB side.

[0213] 3. The backend integration program calls the deep learning algorithm to recognize the captured images and confirm whether the fish body and the standard reference are recognized.

[0214] 4. If the position of the fish body is not placed accurately, the backend integration program will prompt an error message, such as information on adjusting the angle of the fish body.

[0215] 5. When the fish body is placed accurately, the backend integration program will redraw the image and give a green prompt.

[0216] 6. After the backend measurement is completed, the specific test indicators are displayed on the WEB page, such as indicators like body length, body height, head length, fish snout length, caudal peduncle length, caudal peduncle width, tail length, weight, etc.

[0217] The specific operation steps are as follows:

[0218] 1. Place the fish body to be tested on the fish body test board, and the program will display the real-time image.

[0219] 2. At this time, the program software will give a prompt according to the real-time captured image. If it shows green, it means that the fish body is placed correctly and the test is successful.

[0220] The program software will display the specific fish body indicators.

[0221] The final effect diagram of the combined image auxiliary device and the integrated software.

[0222] During the test process, by comparing the results of manual measurement and high-throughput automated measurement, the present invention found that the errors of the two methods in indicators such as body length, body height, body width, and snout length are generally less than 2%, showing the high precision of the new solution. However, in indicators such as caudal peduncle length and caudal peduncle width, the errors are slightly larger. The main reason is that the measurement methods for the caudal peduncle area in the traditional manual measurement method are inconsistent, while the automated measurement solution uses a standardized algorithm for calculation, resulting in a certain difference in measurement standards between the two. In addition, during the manual measurement process, the operator's subjective judgment, different measurement angles, and inconsistent fish body placement methods will all affect the measurement results, while the automated measurement system can provide a more consistent measurement standard, reduce human errors, and make the measurement data more stable and reliable.

[0223] In terms of time efficiency, it takes an average of about 120 seconds to manually measure a fish, while the new solution only takes 1 second to complete the measurement. The total time for measuring 30 fish is reduced from 60 minutes to 30 seconds, which is more than 120 times more efficient. This significant improvement in efficiency means that in large-scale scientific research experiments or commercial application scenarios, researchers can complete more measurement tasks in the same time, greatly improving data collection efficiency. In addition, traditional methods require researchers to repeatedly operate measurement tools, record data, and perform multiple retests to ensure accuracy. The new solution uses an automated system to collect and store data in a database in real time, greatly reducing the workload of repeated operations, reducing human errors, and improving the convenience and traceability of data storage.

[0224] This experimental test shows that the high-throughput automated measurement solution not only improves the measurement speed, but also optimizes the data storage and management methods. The traditional manual measurement method requires researchers to manually record data and organize and enter it in the subsequent stage. This process is prone to human input errors, data loss or inconsistent formats. The new solution can directly store the measurement data into the database through the automated data acquisition system, and supports data backtracking and batch export (CSV, Excel and other formats), which facilitates subsequent data analysis and sharing. At the same time, the automatic measurement system ensures that all measurement data are recorded according to the same standards, which improves the consistency of the data and is suitable for large-scale fish morphology research and scientific research needs in the fields of genetics and ecology.

[0225] The test results show that the high-throughput automated measurement solution has broad prospects in application fields such as scientific research, fishery management and aquaculture monitoring. Its fast, efficient and low-error characteristics make it an important tool for large-scale fish morphology determination, while reducing the labor intensity of experimenters and improving the credibility of data. Future optimization directions include: further improving the measurement algorithm of the caudal peduncle area to make it more in line with the habit of manual measurement and improve the uniformity of measurement standards; adding intelligent calibration functions to the system to enable it to adapt to fish of different sizes and species and improve the adaptability of the system. In addition, by enhancing the ambient light compensation technology, the automated measurement system can still maintain high-precision measurement capabilities under different lighting conditions, so that it can be applied to more complex experiments or natural environment monitoring.

[0226]

[0227]

[0228] The results of this test indicate that the high-throughput automated measurement solution has reached high standards in terms of accuracy. The errors of all measurement indicators are controlled within 2%. Especially for key morphological indicators such as body length, body height, body width, and snout length, the new solution shows high stability and consistency. Compared with the traditional manual measurement method, the new solution eliminates errors caused by inconsistent manual operations, measurement tool deviations, and operator fatigue through standardized image processing algorithms and automated calculation processes, making the measurement data more reliable. The high repeatability accuracy of the automated measurement system ensures the comparability of experimental results and the scientific nature of the data, providing strong technical support for large-scale fish morphological analysis and scientific research data accumulation.

[0229] In terms of measurement efficiency, the new solution shows overwhelming advantages. It only takes 1 second to complete the measurement of a single fish, compared with 120 seconds for the traditional manual measurement, with an efficiency improvement of 120 times, greatly shortening the data acquisition cycle and making the determination of large-scale fish samples more feasible. At the same time, the operation process of the new solution is simple and intuitive. Only by placing the fish body on the measurement platform, the system can automatically complete data acquisition, calculation, and storage, avoiding cumbersome manual measurement steps. This not only reduces the work intensity of the operators but also reduces errors caused by human factors, enabling even non-professional personnel to quickly get started. In addition, this solution eliminates the fatigue effect caused by long-term measurement by personnel, improves the coherence and repeatability of the experiment, and provides an efficient, accurate, and easy-to-operate solution for high-throughput and large-scale fish morphology determination.

[0230] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention shall be covered within the protection scope of the present invention.

Claims

1. A method for high-throughput morphometric determination of fish, characterized in that: The following steps are involved: (1) Collecting fish body images through the image shooting auxiliary component and introducing standard size reference objects; (2) Processing the collected fish images using image recognition software, including data preprocessing, feature extraction, target detection, and morphological measurement; (3) Construct a morphometric model based on a deep learning algorithm to calculate and analyze the morphological data of fish bodies; (4) Store, calculate and output measurement data through back-end integration programs; (5) Display the morphological measurement results on the WEB side and support real-time image acquisition and adjustment prompts.

2. The method for high-throughput morphometric analysis of fish as claimed in claim 1, wherein: The image capture auxiliary component comprises: (1) An integrated fish test background plate, which includes a special background color and a standard reference object; (2) Weight measurement sensor single board computer, used for measuring fish weight; (3) Stable and adjustable support rod for camera fixation and height adjustment; (4) High-definition camera with adjustable parameters, including exposure time and focal length; (5) Bottom plate structure, used to fix support rods and other components; (6) Fill lighting equipment and connectors.

3. The method for high-throughput morphometric analysis of fish as claimed in claim 1, wherein: The image recognition software runs on a computing platform that meets the following hardware configuration: (1) Windows 10 operating system; (2) Intel i5-13500H or above CPU; (3) Independent graphics card GPU; (4) 8 GB or more memory; (5) SSD storage space of 512GB or above.

4. The method for high-throughput morphometric analysis of fish as claimed in claim 1, wherein: The image recognition software includes: (1) Object detection algorithm based on deep learning; (2) Datasets for training object detection models; (3) Real-time image detection and processing module; (4) Back-end integration program, which implements data processing and user interaction based on the Python Flask framework.

5. The method for high-throughput morphometric determination of fishes as claimed in claim 4, characterized in that: The target detection algorithm adopts a fully convolutional neural network architecture and includes the following computational steps: (1) Divide the input image into grids, where each grid is used to detect the target area; (2) Calculate multiple bounding boxes and their confidence scores; (3) predict the conditional probability of the target category; (4) Calculate the final target recognition result based on the confidence score and category probability.

6. The method for high-throughput morphometric determination of fishes as claimed in claim 5, characterized in that: The back-end integration program includes a web service for processing user test requests. The specific process of the web service includes: (1) The user submits a request for fish morphology measurement on the WEB interface; (2) The backend calls the camera to collect real-time images and returns them to the WEB interface; (3) The backend uses deep learning algorithms to detect targets and identify fish and standard reference objects; (4) If the fish body deviates from the position, the system will give an adjustment prompt; (5) When the fish body is placed correctly, the back-end automatically calculates the morphological measurement results and displays them on the WEB interface.

7. The method for high-throughput morphometric determination of fishes as claimed in claim 6, characterized in that: The WEB interface measurement process includes the following steps: (1) The user places the fish on the fish test background plate; (2) The system collects real-time images and determines whether the fish position meets the measurement standards; (3) When the fish body is in the correct position, the system automatically calculates the morphological data such as body length, body length, head length, snout length, caudal peduncle length, caudal peduncle width, tail length and weight; (4) The measurement results are stored in the database, and data export and remote access are supported.

8. A high-throughput morphometric system for fish, characterized in that: include: (1) An image acquisition module, which is used to capture fish images using a high-definition camera and introduce standard size reference objects for scale calibration; (2) Data processing module, including deep learning target detection algorithm, feature extraction algorithm and morphological calculation model, to analyze the collected images and calculate the fish body morphological parameters; (3) The backend integration module, based on the Python Flask framework, is responsible for data storage, morphological calculation, and user interaction; (4) Web interaction module, used to receive user measurement requests, return measurement results, and provide real-time adjustment prompts; (5) Hardware support module, including fish body test background board, adjustable high-definition camera, weight measurement sensor single-board computer, stable support rod and fill light equipment.

9. The fish high-throughput morphometric system according to claim 1, wherein: The image acquisition module comprises: (1) An integrated fish test background plate, including a specific color background and a standard reference object; (2) Adjustable high-definition camera that supports adjustment of exposure time and focal length; (3) A stable support rod with adjustable height for fixing the camera; (4) a weight measurement sensor single-board computer, used to measure the weight of the fish; (5) Fill-in lighting equipment to improve measurement accuracy in low-light environments.

10. The fish high-throughput morphometric system according to claim 1, wherein: The data processing module comprises: (1) Deep learning-based target detection algorithm, using a fully convolutional neural network architecture; (2) Data training set for fish morphology measurement; (3) morphological calculation model, which extracts features from fish body images and calculates morphological parameters such as body length, body length, head length, snout length, caudal peduncle length, caudal peduncle width, tail length, and body weight; (4) Data storage and management module, supporting database storage, measurement record backtracking and data export; (5) The back-end interactive program is integrated into the WEB server to support users to access the measurement data remotely.

Images