Automatic fishery resource evaluation system based on visual AI and operation method thereof

Abstract

The invention discloses a visual AI-based automatic fishery resource evaluation system and an operation method thereof. The system comprises a data acquisition module, a cloud data storage module, an AI identification module, a statistical analysis module and a resource evaluation result module. After the equipment is adopted, the camera is used for collecting data in real time, and high-precision yield estimation is achieved through the visual AI model. The yield authenticity is verified through area calculation, the overall yield is calculated in combination with the transmission hairtail fingerling proportion, and finally automatic resource evaluation and sustainable fishery management are achieved. The system can effectively reduce manual errors, improve data acquisition efficiency, provide a scientific basis for fishery resource protection, and meet the requirements of fishery resource evaluation research on reliability.

Description

Technical Field

[0001] This invention relates to an automated fishery resource assessment system based on visual AI and its operation method, belonging to the field of fishery resource assessment. Background Technology

[0002] Distant-water fisheries are an important component of modern agriculture and the marine economy, and a crucial tool for China to participate in global ocean governance and expand its blue economy. As one of the world's major distant-water fishing nations, my country's distant-water fisheries have maintained steady development in recent years. In 2024, the total output of distant-water fisheries reached 2.1891 million tons, accounting for approximately 15% of the country's total marine catch, with a total output value of 26.703 billion yuan, demonstrating strong production capacity and economic value. Currently, my country has approximately 2,470 distant-water fishing vessels, directly employing about 100,000 people. Their operations cover the world's three major oceans and polar waters, while also conducting legal and compliant fishing activities within the exclusive economic zones of more than 30 countries and regions. my country has joined eight regional fisheries management organizations, covering the management organizations corresponding to the world's major distant-water fishing areas, achieving 100% participation. In 2023, the white paper "Development of China's Distant-Water Fisheries" pointed out the need to earnestly fulfill international conventions, conserve international fishery resources, continuously improve the collection and reporting of basic production statistics for distant-water fisheries, improve data quality, and promote data sharing and integrated management. A comprehensive data collection system for distant-water fisheries has been established, covering information on distant-water fishing enterprises, distant-water fishing vessels, vessel position monitoring, fishing logs, catch transshipment, national observers, port sampling, scientific surveys, and productive exploration fishing. In accordance with the regulations of relevant regional fisheries management organizations, various types of fisheries data are submitted in a timely manner.

[0003] However, traditional yield estimation methods have some limitations. For example, manually recorded yields are easily affected by various factors, including human error, and in complex fisheries environments, data acquisition is often cumbersome, time-consuming, and has low accuracy. These factors not only affect the scientific assessment of fishery resources but also pose significant challenges to fisheries management departments in implementing catch quotas and resource protection policies.

[0004] With the advancement of intelligent technologies, automated detection methods based on computer vision and deep learning have gradually become powerful tools for improving the accuracy and efficiency of fisheries production data. While existing visual recognition technologies can automatically acquire images through cameras and perform a certain degree of fish species identification, they still face challenges in fish species identification in the Northwest Pacific light-purse seine fisheries. When calculating the proportion of major catch species, Japanese mackerel and Far Eastern sardines overlap or partially obscure each other, causing existing algorithms to frequently produce errors in identifying individual numbers and estimating fish weight, leading to inaccurate yield estimates.

[0005] Scientific fisheries resource assessment is crucial for determining fisheries quotas and effectively managing fish populations. Data accuracy is paramount in this assessment; erroneous data can lead to misjudgments of the Japanese mackerel resource status, consequently impacting the scientific rigor and sustainability of distant-water fisheries compliance and management decisions. Therefore, there is an urgent need to develop a reliable data collection method to ensure high accuracy and consistency in the acquired fish yield information. Based on this, advanced visual AI technology can be used to precisely process and analyze fisheries data, effectively reducing human error and enhancing the scientific validity and credibility of resource assessments, thus providing robust data support for fisheries resource management and sustainable development. Summary of the Invention

[0006] To address the aforementioned technical issues, this invention provides an automated fishery resource assessment system and its operation method based on visual AI, integrating automated fish species identification, reliable yield data collection, and accurate resource assessment. The system collects data via a camera and uploads it to the cloud for AI visual recognition, thereby improving the accuracy of yield calculation and the reliability of fish species ratio identification. This effectively facilitates the automation and intelligentization of fishery resource assessment and enhances the reliability of assessment results.

[0007] An automated fishery resource assessment system based on visual AI includes a data acquisition module, a cloud data storage module, an AI recognition module, a statistical analysis module, and a resource assessment result module.

[0008] In a preferred embodiment, the data acquisition module is responsible for acquiring camera images and preprocessing the data, and storing them in the cloud data storage module via the ship's wireless network.

[0009] As a preferred embodiment, the AI ​​recognition module develops a visual AI model based on a convolutional attention module (parallel multi-convolution (1×1, 3×3, 5×5, 7×7) and channel-spatial attention) to enhance feature extraction capabilities, improve the accuracy and robustness of fish species recognition, and achieve automated recognition of the authenticity of fish species ratio and yield.

[0010] In a preferred embodiment, the statistical analysis module calculates the yield share of the research species by analyzing the proportion of the research species on the conveyor belt. The formula for calculating the yield of Japanese mackerel is as follows:

[0011] In the formula, t is the image recognition time, and N is the total recognition time. The proportion of the species studied at time t.

[0012] In a preferred embodiment, the resource assessment result module divides the collected yield data into catch time series at different time scales, and dynamically adjusts the training data window length of the Catch-Maximum Sustainable Yield (Catch-MSY) model in conjunction with the real-time catch data of the research species collected by the camera, so as to achieve automated resource assessment.

[0013] An operational method for an automated fishery resource assessment system based on visual AI, specifically including the following operational steps: (1) Data collection and uploading: The shipborne camera automatically collects images of the catch on the conveyor belt and uploads them to the cloud data storage module via the ship’s wireless network.

[0014] (2) AI recognition and statistical analysis: The cloud-based visual AI model automatically identifies various fish species, statistically studies the proportion of catches on the conveyor belt, and calculates the corresponding catch percentage.

[0015] (3) Automated resource assessment and result output: The yield data is divided into catch time series at different time scales, and the training data window of the Catch-MSY model is dynamically adjusted in combination with the real-time identification results to automatically generate fishery resource assessment results and fishery production strategies.

[0016] Beneficial Effects: By employing the equipment described in this application, real-time data collection via cameras and high-precision yield estimation through a visual AI model are achieved. Yield accuracy is verified through area calculations, and overall yield is calculated by combining the proportion of fish species in the conveyor belt, ultimately realizing automated resource assessment and sustainable fisheries management. This system effectively reduces human error, improves data collection efficiency, and provides a scientific basis for fisheries resource protection, meeting the reliability requirements of fisheries resource assessment research. Attached Figure Description

[0017] Figure 1 This is a cross-sectional view of the automated fishery resource assessment device of the present invention.

[0018] Figure 2 This is a schematic diagram of sample images in the data acquisition module of this invention. Detailed Implementation

[0019] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings: These embodiments are implemented based on the technical solution of the present invention, and provide detailed implementation methods and specific operation processes, but the protection scope of the present invention is not limited to the following embodiments.

[0020] like Figure 1As shown, an automated fishery resource assessment system based on visual AI includes a data acquisition module (1), a cloud data storage module (2), an AI recognition module (3), a statistical analysis module (4), and a resource assessment result module (5).

[0021] Based on this implementation, a typical usage is as follows: The automated fishery resource assessment equipment based on visual AI has a data acquisition module (1). The data acquisition module (1) is responsible for camera image acquisition and data preprocessing, and transmits the data to the cloud data storage module (2) via the ship's wireless network. Cameras are deployed on all light-filled purse seine fishing vessels in the Northwest Pacific to collect fishery production data from March to December each year. Stable data acquisition and transmission are achieved through technical integration with the cooperating camera manufacturer. The collected data is aggregated via the ship's wireless network and uploaded to a cloud server located in Shanghai to ensure the real-time performance and integrity of data transmission.

[0022] like Figure 2 As shown, the cloud data storage module (2) is responsible for storing and managing a large amount of high-resolution image data, providing high-speed data reading and access capabilities. It adopts a distributed storage architecture to ensure data security and scalability, and provides data support for the subsequent AI recognition module (3) and statistical analysis module (4).

[0023] The AI ​​recognition module (3) is used to identify the proportion of Japanese mackerel and sardines in different months and spatial scales in 2024 stored in the cloud data storage module (2), identify and calculate the yield, and verify the authenticity of the corresponding fishing logs in time and space, so as to accumulate real and reliable data for automated resource assessment.

[0024] The statistical analysis module (4) analyzed the catch data obtained by the AI ​​recognition module (3) at different time scales and found that the proportion of sardines was significantly higher in the early fishing season, and its yield only gradually increased after the arrival of the Japanese mackerel fishing season. This trend reveals the temporal distribution characteristics of different fish species and provides a scientific basis for the resource assessment results module (5).

[0025] The resource assessment results module (5) divides the collected yield data into catch time series at different time scales, and dynamically adjusts the training data window length of the Catch-Maximum Sustainable Yield (Catch-MSY) model in conjunction with the real-time Japanese mackerel yield data collected by the camera. The resource assessment results show that the estimation results of MSY value are more sensitive to the time series length of the catch data.

[0026] An operational method for an automated fishery resource assessment system based on visual AI, specifically including the following operational steps: (1) Data collection and uploading: The shipborne camera automatically collects images of the catch on the conveyor belt and uploads them to the cloud data storage module via the ship’s wireless network.

[0027] (2) AI recognition and statistical analysis: The cloud-based visual AI model automatically identifies various fish species, statistically studies the proportion of catches on the conveyor belt, and calculates the corresponding catch percentage.

[0028] (3) Automated resource assessment and result output: The yield data is divided into catch time series at different time scales, and the training data window of the Catch-MSY model is dynamically adjusted in combination with the real-time identification results to automatically generate fishery resource assessment results and fishery production strategies.

[0029] This invention can effectively reduce human error, improve data collection efficiency, and provide a scientific basis for fishery resource protection, meeting the reliability requirements of fishery resource assessment research.

[0030] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. An automated fishery resource assessment system based on visual AI, characterized in that, It includes a data acquisition module, a cloud data storage module, an AI recognition module, a statistical analysis module, and a resource assessment results module.

2. The automated fishery resource assessment system based on visual AI according to claim 1, characterized in that, The data acquisition module is responsible for acquiring camera images and preprocessing the data, and then storing them in the cloud data storage module via the ship's wireless network.

3. The automated fishery resource assessment system based on visual AI according to claim 1, characterized in that, The AI ​​recognition module develops a visual AI model based on a convolutional attention module to enhance feature extraction capabilities, improve the accuracy and robustness of fish species identification, and achieve automated identification of fish species ratios and yield authenticity.

4. The automated fishery resource assessment system based on visual AI according to claim 1, characterized in that, The statistical analysis module calculates the yield share of the research species by analyzing the proportion of the research species on the conveyor belt. The formula for calculating the yield of Japanese mackerel is as follows: ， In the formula, t is the image recognition time, and N is the total recognition time. The proportion of the species studied at time t.

5. The automated fishery resource assessment system based on visual AI according to claim 1, characterized in that, The resource assessment results module divides the collected yield data into catch time series at different time scales, and dynamically adjusts the training data window length of the catch-maximum sustainable yield model in combination with the real-time harvesting data of the research species collected by the camera, so as to achieve automated resource assessment.

6. An operational method for an automated fishery resource assessment system based on visual AI, characterized in that, The specific steps include the following: (1) Data collection and uploading: The shipborne camera automatically collects images of the catch on the conveyor belt and uploads them to the cloud data storage module via the ship’s wireless network; (2) AI recognition and statistical analysis: The cloud-based visual AI model automatically identifies various fish species, statistically studies the proportion of the catch on the conveyor belt, and calculates the corresponding catch percentage; (3) Automated resource assessment and result output: The production data is divided into catch time series at different time scales, and the training data window of the Catch-MSY model is dynamically adjusted in combination with the real-time recognition results to automatically generate fishery resource assessment results and fishery production strategies.

Images