A fishing effort estimation method, fishery resource distribution detection method and system

Abstract

The present invention discloses a fishing effort estimation method, a fishery resource distribution detection method and system, and relates to the technical field of fishery management. The present invention includes constructing an SMA‑lightGBM model, and using a ship position feature vector for training to obtain a fishing behavior recognition model; obtaining the ship position data of fishing vessels in a target sea area during a target period; pre-processing and extracting the feature vector of the ship position data of fishing vessels in the target sea area during the target period, and then inputting the data into a fishing behavior recognition model to identify the fishing behavior and time period distribution of fishing vessels in each fishing area in the target sea area during the target period; accumulating the fishing behavior and time period distribution of fishing vessels in each fishing area in the target sea area during the target period to obtain the fishing effort of each fishing area in the target sea area during the target period. The present invention can timely and automatically estimate the fishing effort of different areas in the sea area, and detect the distribution of fishery resources in the target sea area.

Description

Technical Field

[0001] The present invention belongs to the technical field of fishery management, and in particular relates to a fishing effort estimation method, a fishery resource distribution detection method and a system. Background Art

[0002] With the continuous development and utilization of global fishery resources, many fishery resources are facing the threat of overfishing, and the sustainable management of fishery resources has become particularly important. Fishing effort is one of the key indicators to measure the intensity of fishery activities. A reasonable estimate of fishing effort can help fishery managers allocate fishery resources more effectively, avoid resource depletion, and promote the long-term sustainable development of fisheries.

[0003] At present, the estimation of fishing effort usually relies on manually recorded data such as the number of times a fishing vessel goes out to sea, the operating time, and the type of fishing gear. However, due to the complexity of fishing vessel operations, the diversity of fishing behaviors, and the incompleteness of data records, traditional fishing effort estimation methods have problems such as low accuracy and poor real-time performance. This leads to inaccurate results of fishery resource distribution detection, which further affects the effective decision-making of fishery management departments on resource development intensity and protection measures. Summary of the invention

[0004] The purpose of the present invention is to provide a fishing effort estimation method, a fishery resource distribution detection method and system, which can automatically and timely estimate the fishing effort in different areas of the sea area and detect the distribution of fishery resources in the target sea area by collecting the position data of fishing vessels for analysis.

[0005] In order to solve the above technical problems, the present invention is achieved through the following technical solutions: The present invention provides a method for estimating fishing effort, comprising: Obtain the position data of multiple fishing vessels during the sample period; Preprocessing and extracting feature vectors of all the ship position data to obtain feature vectors of all the ship position data as ship position feature vectors; Constructing an SMA-lightGBM model, and using the ship position feature vector for training to obtain a fishing behavior recognition model; Obtain the position data of fishing vessels in the target sea area during the target period; The position data of fishing boats in the target sea area during the target period are pre-processed and feature vectors are extracted, and then input into the fishing behavior recognition model to identify the fishing behavior and time distribution of fishing boats in each fishing area of ​​the target sea area during the target period; The fishing behavior and time distribution of fishing boats in each fishing area in the target sea area during the target period are accumulated to obtain the fishing effort of each fishing area in the target sea area during the target period.

[0006] The present invention also discloses a method for detecting the distribution of fishery resources, comprising: Obtain the fishing effort of each fishing area in the target sea area during the target period; The ratio of the fishing efforts in each fishing area of ​​the target sea area during the target period is taken as the distribution ratio of fishery resources in each fishing area of ​​the target sea area.

[0007] The present invention also discloses a fishery resource distribution detection system, comprising: A receiving unit, used for receiving the position data of multiple fishing vessels within a sample period; A fishing calculation unit, used to obtain the position data of multiple fishing vessels during the sample period; Preprocessing and extracting feature vectors of all the ship position data to obtain feature vectors of all the ship position data as ship position feature vectors; Constructing an SMA-lightGBM model, and using the ship position feature vector for training to obtain a fishing behavior recognition model; Obtain the position data of fishing vessels in the target sea area during the target period; The position data of fishing boats in the target sea area during the target period are pre-processed and feature vectors are extracted, and then input into the fishing behavior recognition model to identify the fishing behavior and time distribution of fishing boats in each fishing area of ​​the target sea area during the target period; The fishing behavior and time distribution of fishing boats in each fishing area of ​​the target sea area during the target period are accumulated to obtain the fishing effort of each fishing area in the target sea area during the target period; Fishery resource detection unit, used to obtain the fishing effort of each fishing area in the target sea area during the target period; The ratio of the fishing efforts in each fishing area of ​​the target sea area during the target period is taken as the distribution ratio of fishery resources in each fishing area of ​​the target sea area.

[0008] The present invention collects the position data of fishing boats through a receiving unit, and then receives the position data of fishing boats through a fishing calculation unit and constructs an SMA-lightGBM model for training, thereby obtaining a fishing behavior recognition model, and identifying the fishing behavior of fishing boats based on it, and finally accumulating the fishing effort of the target sea area. In this process, it is only necessary to automatically collect and analyze the position data of fishing boats, and the fishing effort of different areas in the sea area can be automatically estimated in time, and the distribution of fishery resources in the target sea area can be obtained based on the fishing effort of different areas.

[0009] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0011] Figure 1 A schematic diagram of functional units and information flow of a fishery resource distribution detection system according to an embodiment of the present invention; Figure 2 A schematic diagram of a step flow chart of a fishing calculation unit according to an embodiment of the present invention; Figure 3 A schematic diagram of the steps of the fishery resource detection unit in one embodiment of the present invention; Figure 4 This is a schematic diagram of the step flow of step S3 in one embodiment of the present invention; Figure 5 This is a schematic diagram of the step flow of step S33 in one embodiment of the present invention; Figure 6 This is a schematic diagram of a step flow chart of step S332 in an embodiment of the present invention; Figure 7 This is a schematic diagram of the step flow of step S5 of the present invention in one embodiment; Figure 8 A schematic diagram of a step flow chart of step S53 in an embodiment of the present invention; Fig. 9 This is a schematic diagram of a step flow chart of step S534 in an embodiment of the present invention; Fig.10 This is a schematic diagram of the step flow of step S6 in one embodiment of the present invention; In the accompanying drawings, the components represented by the reference numerals are listed as follows: 1- Receiving unit, 2- Fishing calculation unit, 3- Fishery resources detection unit. DETAILED DESCRIPTION

[0012] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below in conjunction with the accompanying drawings.

[0013] It should be noted that the terms "first", "second", etc. in this application are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0014] See also Figure 1 As shown, the present invention provides a fishery resource distribution detection system, which includes a receiving unit 1, a fishing calculation unit 2 and a fishery resource detection unit 3 in terms of functional units. The receiving unit 1 is used to obtain various parameters of the fishing boat, and the fishing calculation unit 2 and the fishery resource detection unit 3 are virtual functional units, which can be computing devices in specific applications.

[0015] The receiving unit 1 is used to first execute step S0 to obtain the position data of multiple fishing vessels in the sample period through the ship position database. Of course, the position data of fishing vessels in the sample period can also be extracted through the ship position collection system. The ship tracking device carried by the fishing vessel can obtain real-time and accurate key information such as the longitude and latitude, heading and speed of the fishing vessel. Among them, VMS and AIS, as the most fertile data sources, monitor the activity tracks of most fishing vessels in the world, which makes it possible to mine the operating status of fishing vessels from the tracks, thereby evaluating the fishing effort. Because the speed of the operating status of the fishing vessel is affected by many factors such as water flow, tide, weather, net size, target fish species and captain's preference, and the speed alone has the disadvantage of characteristic limitations, this makes the threshold division method difficult to apply to all fishing vessels and the estimated value of fishing effort is too high.

[0016] Due to the above problems, the fishing calculation unit 2 in this scheme is limited to executing step S1 to obtain the position data of multiple fishing boats during the sample period. The position data used in this scheme comes from the National Fishing Vessel Dynamic Monitoring and Management System (https: / / www.vmscenter.com), which contains the Beidou position data of trawlers in Liaoning, Hebei, Shandong and Jiangsu provinces along the Yellow Sea and Bohai Sea. The time coverage of the data is from 0:00 on September 1, 2019 to 24:00 on December 31, 2021. A total of 4917 trawlers and 195,020,382 position data were collected. The fishing vessel position data record includes attributes such as ship name, MMSI code (fishing vessel water mobile communication service identification code), longitude, latitude, speed, heading, time, etc. In the preprocessing process of the position data, firstly, according to the overall range of the data, the data with missing items, missing items and over-limit in the longitude, latitude, speed and time in the position report are eliminated. Calculate the distance from each ship position to the coastline, and divide the data segments according to the time of entry and exit. Since the satellite may be affected by signal fluctuations during the process of receiving the returned data, and occasionally there may be situations where the transmitting source is blocked by humans, if the time interval between the two consecutive position reports exceeds 3 hours, the trajectory segment will be split, and finally the trajectory segment with less than 5 points will be deleted.

[0017] The position data of 33 trawlers from September 2019 to December 2021 were randomly selected from the collected data of trawlers in the Yellow Sea and Bohai Sea, totaling 788,908 positions. The position data were marked according to the divided segments, and the position and speed of each segment were used to draw a sea route map based on time in the geographic information system software. The acquired fishery information and prior knowledge such as the characteristics of trawling fishing operations, expert experience, and fishermen interviews were combined with the fishing time recorded in the fishing log to comprehensively analyze the operation behavior in the track and record the start and end time of the operation, that is, the position in the fishing state time period was marked as 1, and the non-fishing state was marked as 0. During the calibration process, this study adopted the method of multi-person calibration and replacement of personnel for review to ensure that the calibration results are objective and correct.

[0018] Next, step S2 can be executed to pre-process all the ship position data and extract feature vectors to obtain the feature vectors of all the ship position data as the ship position feature vector. In order to identify fishing behavior, this scheme extracts 11-dimensional features from each ship position data to form a feature vector. First, this study downloaded the coastline on Natural Earth and calculated the distance between the ship position and the coastline in each data. According to the fishing boat berthing information and high-definition remote sensing maps, the ports along the Yellow Sea and Bohai Sea coast of China were manually marked, and the distance from the ship position to the nearest port in each data was calculated. At the same time, the time difference before and after the position reporting data, the actual distance, the theoretical average speed, the azimuth difference, the azimuth change rate were calculated, and the real-time speed, local time and month were also recorded.

[0019] Next, step S3 can be executed to build the SMA-lightGBM model, and the ship position feature vector can be used for training to obtain a fishing behavior recognition model. In this solution, Beidou ship position data is used as the basis, and data features are extracted and feature vectors are constructed through feature engineering, and the lightGBM algorithm optimized by SMA is used to determine whether a fishing vessel is engaged in fishing.

[0020] Next, step S4 can be executed to obtain the position data of fishing boats in the target sea area during the target period. Next, step S5 can be executed to pre-process the position data of fishing boats in the target sea area during the target period and extract the feature vector, and then input it into the fishing behavior recognition model to identify the fishing behavior and time distribution of fishing boats in each fishing area in the target sea area during the target period. Finally, step S6 can be executed to accumulate the fishing behavior and time distribution of fishing boats in each fishing area in the target sea area during the target period to obtain the fishing effort of each fishing area in the target sea area during the target period.

[0021] Fishery resources in the ocean are usually highly correlated with fishing effort. In order to detect the distribution of fishery resources in the target sea area, the present solution can be used by the fishery resource detection unit 3 to perform step S031 to obtain the fishing effort of each fishing area in the target sea area during the target period. Then, step S032 can be performed to use the ratio of the fishing effort of each fishing area in the target sea area during the target period as the distribution ratio of fishery resources in each fishing area in the target sea area. Of course, in actual applications, different sea areas can add corrections based on the fishing effort.

[0022] See also Figure 4As shown, in the process of constructing the SMA-lightGBM model and training it, the SMA algorithm optimizes the lightGBM hyperparameters according to the calculation results of the fitness function until the optimal parameter group is found. The entire algorithm is mainly divided into: data collection, cleaning and annotation, feature vector extraction, and model training. During the training process, SMA is used to optimize the lightGBM parameter optimization process, and then the model is tested using external data to obtain model evaluation indicators. Specifically, the restricted step S31 can be executed to divide all the ship position feature vectors into training sets and test sets. Next, step S32 can be executed to randomly and evenly divide the training set into multiple subsets. Next, step S33 can be executed to use one of the subsets as a validation set and the remaining subsets as training sets. Next, step S34 can be executed to use each subset as a validation set in turn, and the remaining subsets as training sets, and the SMA-lightGBM model can be trained and verified to obtain the SMA-lightGBM model with the optimal parameters as a fishing behavior recognition model. Finally, step S35 can be executed to test the fishing behavior recognition model using the ship position feature vectors in the test vector set to obtain average accuracy, sensitivity, specificity and Matthews correlation coefficient as generalization ability evaluation indicators of the fishing behavior recognition model. Next, each step will be described in detail.

[0023] See also Figure 5 As shown, in the process of training the SMA-lightGBM model, in order to improve the training rate, firstly, step S331 can be executed to use each subset as a validation set and the remaining subsets as training sets to obtain multiple training set combinations. Next, step S332 can be executed to continuously adjust the SMA parameters for each training set combination to train and verify the SMA-lightGBM model to obtain the average accuracy. Finally, step S333 can be executed to use the SMA-lightGBM model with the highest average accuracy as the SMA-lightGBM model with the optimal parameters. Specifically, this scheme overcomes the limitations of previous models by establishing an SMA-optimized intelligent model, which can identify the fishing behavior of trawlers faster and more accurately. This scheme can use lightGBM as a fishing behavior classifier for fishing vessels, which is a high-level parallel gradient boosting framework. It introduces a histogram algorithm based on XGBoost, proposes a unilateral gradient sampling and mutually exclusive feature bundling algorithm for data processing optimization, abandons the decision tree growth strategy of layer-by-layer growth, and adopts a leaf growth strategy with depth restrictions, which ensures high efficiency while preventing overfitting.

[0024] See also Figure 6As shown, for each training set combination, in the process of adjusting the SMA parameters to train and verify the SMA-lightGBM model, step S3321 can be first executed to initialize the SMA parameters. Next, step S3322 can be executed to input the ship position feature vector in the training set into the SMA-lightGBM model for training. Next, step S3323 can be executed to verify the trained lightGBM model using the ship position feature vector in the verification set to obtain the accuracy of the SMA-lightGBM model and its changing trend after the SMA parameters are optimized. Next, step S3324 can be executed to continuously adjust the population size (popsize) and the maximum number of iterations (Max Miter) in the SMA parameters until the SMA-lightGBM model reaches the target accuracy. Finally, step S3325 can be executed to calculate and obtain the average accuracy of the SMA-lightGBM model to achieve the target accuracy after the SMA parameters have been adjusted.

[0025] Specifically, the histogram algorithm used in the above scheme uses the histogram algorithm to replace the traditional pre-sorting algorithm, solving the huge memory consumption and low efficiency of XGBoost when processing large-scale high-dimensional data. Convert continuous floating-point data to bin data. When traversing the data, calculate the cumulative value of each discrete value in the histogram, and determine the optimal segmentation point based on the distribution of discrete values. The unilateral gradient sampling in this scheme can reduce the number of data instances while maintaining the accuracy of the decision tree. LightGBM proposes unilateral gradient sampling. In order to further improve the efficiency of training calculations, this scheme can also use mutually exclusive feature bundling. LightGBM not only downsamples the data, but also samples the feature dimensions, which further improves the model training speed. The EFB algorithm binds mutually exclusive features together to reduce the dimension. EFB is an NP-hard problem, that is, a small number of sample points are allowed between features that are not mutually exclusive. Allowing a small part of the conflict can obtain a smaller number of feature bundles, further improving the calculation efficiency.

[0026] In this scheme, SMA optimization uses the slime mold algorithm to optimize lightGBM. The slime mold algorithm is a novel meta-heuristic intelligent optimization algorithm that simulates the diffusion and lost behavior of slime molds. It uses adaptive weights to simulate the process of positive and negative feedback generated by the "slime mold propagation wave" based on the biological oscillator, forming an optimal food connection path with good exploration ability and development tendency. It is an effective optimization algorithm. Slime molds connect multiple food sources and establish a network of veins of varying thickness between food sources. When the veins are close to the food source, the signals emitted by the slime mold stimulate the internal biological oscillator to generate propagation waves, promoting the cytoplasm flow of the veins. The higher the food concentration, the stronger the wave signal, the faster the cytoplasm flows, and the thicker the veins, and vice versa.

[0027] In the above steps, lightGBM is used to classify the fishing behavior of trawlers. However, since the use of LightGBM involves adjusting multiple parameters and the parameter values ​​have a great impact on the model classification effect, grid search and manual parameter adjustment are time-consuming and laborious. Therefore, we propose to use SMA to optimize lightGBM. In the process of optimizing lightGBM parameters with the SMA algorithm, the number and range of parameters to be optimized are first determined. The parameters to be optimized in lightGBM are set as the number of leaf nodes of each decision tree, the depth of the tree, the minimum number of leaf node samples, regularization parameters γ and λ, the sample sampling rate of the learner and the feature sampling rate, the learning rate and the number of iterations. The upper and lower limit arrays ub and lb of the parameters are set according to the range of the 9 parameters. Then, the average cross-validation accuracy of the lightGBM model is used as the fitness function of SMA. Next, the slime mold population is initialized and the fitness of all slime molds is calculated. Then, it is determined whether the maximum number of iterations has been reached. If the current number of iterations exceeds the maximum number of iterations, the parameter optimization result is output, otherwise the slime mold position is updated and the optimization continues. After the above steps, the trawler fishing behavior recognition model based on SMA-LightGBM is obtained.

[0028] See also Figure 7 As shown, since the fishing boats are dispersed in the sea area, they are not evenly distributed at the same time. In order to divide the key fish-catching areas in the target sea area, step S51 can be first executed to divide the target time period into multiple unit time periods of the same duration. Next, step S52 can be executed to pre-process the position data of the fishing boats in the target sea area in each unit time period and extract the feature vector, and then input it into the fishing behavior recognition model to identify the fishing behavior of the fishing boats in the target sea area in each unit time period. Next, step S53 can be executed to obtain the sea area distribution of each fishing area in the target sea area in the unit time period for each unit time period according to the position data and fishing behavior of the fishing boats in the target sea area. Finally, step S54 can be executed to summarize the fishing behavior and time period distribution of the fishing boats in the sea area distribution range of each fishing area in each unit time period to obtain the fishing behavior and time period distribution of the fishing boats in each fishing area of ​​the target sea area in the target time period.

[0029] See also Figures 8 to 9 As shown, in the process of dividing the fishing area in each unit time period, first, step S531 can be executed to extract the position of the fishing boat in the fishing state according to the position data and fishing behavior of the fishing boat in the target sea area. Then, step S532 can be executed to select several positions from the positions of the fishing boats in the fishing state as reference positions. Then, step S533 can be executed to calculate the distance between each reference position and the fishing boats other than the reference position.

[0030] Then, step S534 can be executed to classify the position of each fishing boat and the reference position with the shortest distance into the same position group. Specifically, step S5341 can be executed first to obtain the geographic center of the distribution range of the positions of all fishing boats in each position group. Next, step S5342 can be executed to use the position of the fishing boat closest to the geographic center as the reference position after iteration. Next, step S5343 can be executed to determine whether the reference position after iteration has changed. If so, it means that the above grouping is insufficient, so next, steps S5341 to S5343 can be executed to continuously calculate the geographic center and iteratively generate the reference position. If not, it means that the above steps have been fully grouped, so next, step S5344 can be executed to obtain several position groups. Finally, step S5345 can be executed to use the reference position and fishing boats in the same position group as the reference position and the fishing boat with position correlation.

[0031] Next, step S535 may be executed to respectively use the position distribution range of fishing boats having position correlation with each reference position as the fishing area for each reference position. Finally, step S536 may be executed to summarize and obtain the sea area distribution of each fishing area in the target sea area within the unit time period.

[0032] In order to provide additional explanation for the implementation process of the above-mentioned steps S531 to S536, source codes of some functional modules are provided, and comparative explanations are provided in the comments.

[0033] import numpy as np from geopy.distance import geodesic # Call the fishing behavior recognition model class FishingBehaviorModel: def predict(self, features): # Simulate fishing behavior identification, 1 indicates fishing behavior, 0 indicates non-fishing behavior return np.random.choice([0, 1], size=features.shape[0]) # Calculate the geographic distance between two points def calculate_distance(pos1, pos2): Use geopy to calculate the geographic distance between two points (unit: nautical miles) return geodesic(pos1, pos2).kilometers # Extract the position of fishing boats in fishing state def extract_fishing_positions(position_data, predictions): Extract the location of fishing boats in fishing state based on fishing behavior fishing_positions = [] for i, pred in enumerate(predictions): if pred == 1: # Only extract fishing boats that are in fishing state fishing_positions.append((position_data[i]["latitude"],position_data[i]["longitude"])) return fishing_positions # Calculate geographic center def calculate_geographic_center(positions): Calculate the geographic center of a set of geographic locations latitudes = [pos[0] for pos in positions] longitudes = [pos[1] for pos in positions] return np.mean(latitudes), np.mean(longitudes) # Position grouping function def group_positions_by_distance(fishing_positions, base_positions): Group fishing vessels in the fishing state according to the shortest distance to the reference position position_groups = {tuple(base): [] for base in base_positions} for pos in fishing_positions: # Find the closest reference position to the current fishing boat closest_base = min(base_positions, key=lambda base: calculate_distance(pos, base)) position_groups[tuple(closest_base)].append(pos) return position_groups # Get the baseline position after iteration def update_base_positions(position_groups): Update the base location to the geographic center of each group new_base_positions = [] for base, group in position_groups.items(): if group: new_base = calculate_geographic_center(group) new_base_positions.append(new_base) return new_base_positions # Determine whether the reference position has changed def base_positions_changed(old_bases, new_bases): Determine whether the reference position has changed return set([tuple(base) for base in old_bases]) != set([tuple(base)for base in new_bases]) # Main function: Extract fishing area def extract_fishing_areas(position_data, predictions): Extract fishing areas based on fishing behavior and location # Step 1: Extract the location of fishing boats in fishing state fishing_positions = extract_fishing_positions(position_data,predictions) # Initialize the reference position and select several positions as the reference position (here several points are randomly selected as the initial reference position) if len(fishing_positions)<= 2: base_positions = fishing_positions # If there are few fishing boats, use the fishing boats as the base position else: base_positions = [fishing_positions[i] for i in np.random.choice(range(len(fishing_positions)), 2, replace=False)] # Iteratively update the base position while True: # Step 2: Group fishing boats by reference position position_groups = group_positions_by_distance(fishing_positions,base_positions) # Step 3: Update the base position to the geographic center of each group new_base_positions = update_base_positions(position_groups) # Step 4: Determine whether the reference position has changed if not base_positions_changed(base_positions, new_base_positions): break # The base position no longer changes, ending the iteration base_positions = new_base_positions # Final reference position and fishing area fishing_areas = [] for base, group in position_groups.items(): if group: center = calculate_geographic_center(group) fishing_areas.append({ "base_position": base, "geographic_center": center, "positions": group }) return fishing_areas # Sample data: Ship position data during the target period position_data = [ {"timestamp": "2024-10-12 10:00:00", "latitude": 30.5, "longitude":120.5}, {"timestamp": "2024-10-12 10:30:00", "latitude": 30.6, "longitude":120.6}, {"timestamp": "2024-10-12 11:00:00", "latitude": 30.7, "longitude":120.7}, {"timestamp": "2024-10-12 11:30:00", "latitude": 30.8, "longitude":120.8} ] # Example feature vector (examples have been extracted from the data) features = np.array([[0.5, 1.2], [0.3, 1.0], [0.8, 0.9], [0.6, 1.1]]) # Fishing behavior recognition model (the example model has been trained) fishing_model = FishingBehaviorModel() # Step 1: Use the fishing behavior recognition model to predict fishing behavior predictions = fishing_model.predict(features) # Step 2: Extract fishing areas fishing_areas = extract_fishing_areas(position_data, predictions) # Output results for area in fishing_areas: print(f"Base position: {area['base_position']}, Geographic center: {area['geographic_center']}, Number of fishing boats: {len(area['positions'])}") The above Python code divides the fishing area in the target sea area through the position data of fishing vessels and the fishing behavior prediction model. First, identify the fishing vessels in the fishing state from the position data, select several reference positions and group them according to the distance. Then, by iteratively updating the reference positions, the fishing vessels in the fishing state are gathered into suitable areas until the reference positions no longer change. Finally, the positions of the fishing vessels are grouped to generate various fishing areas, and the geographic center and number of fishing vessels in each area are output. This process can help effectively identify the distribution of fishing behavior of fishing vessels.

[0034] See also Fig.10 As shown, in the process of calculating the fishing effort, for each fishing area, first, step S61 can be executed to obtain the time proportion of each fishing boat in the target period of time according to the fishing behavior and time distribution of the fishing boat in the target period of time, and then step S62 can be executed to accumulate the time proportion of all fishing boats in the fishing area in the target period of time to obtain the fishing effort of the fishing area in the target period of time. Finally, step S63 can be executed to summarize and obtain the fishing effort of each fishing area in the target sea area in the target period of time.

[0035] From September 2019 to April 2023, the fishing effort was mostly distributed in three areas, including ①: 121~124E, 35.5~39N, with an overall grid coverage of 26.25% of the study area. The total time spent by trawlers searching for fish and fishing in this area accounted for about 55.3% of the total time in 5 years. The fish resources in this area are rich. ②: 119.7~122.7E, 33.8~35.5N, the overall grid coverage was 12.75% of the study area, and the total fishing time accounted for about 18.75%, which is the second largest fishing hotspot area. ③: 123~124E, 33.5~35N, the overall grid coverage was 3.75% of the study area. The total time spent by trawlers searching for fish and fishing in this area accounted for about 6.14%. The area is small but the fishing effort is relatively high.

[0036] In order to further improve the accuracy of fishing effort assessment and reduce estimation errors, this scheme can also be optimized from three perspectives: ① In terms of features, the accuracy of the model can be improved by extracting effective spatial information between ships as feature parameters. ② In terms of algorithms, deep learning networks such as transformers can be used to construct a trawler fishing behavior recognition model to improve model efficiency. ③ In terms of data, since the workload of ship position data annotation is very large, the amount of trawler data selected in this scheme is limited. In order to optimize the model effect, the amount of data can be further expanded. In addition, the power calculation process should be optimized, and the method of calculating the working power of fishing vessels should be further optimized to improve the accuracy of fishing effort estimation.

[0037] The flow chart and block diagram in the accompanying drawings show the possible architecture, function and operation of the device, system, method and computer program product according to multiple embodiments of the present application. In this regard, each square frame in the flow chart or block diagram can represent a part of a module, program segment or instruction, and a part of the module, program segment or instruction includes one or more executable instructions for realizing the logical function of the specification. In some alternative implementations, the functions marked in the square frame can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous square frames can actually be executed substantially in parallel, and they can also be executed in reverse order sometimes, depending on the functions involved.

[0038] It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart, can be implemented by hardware that performs the corresponding function or action, such as a circuit or ASIC (Application Specific Integrated Circuit), or can be implemented by a combination of hardware and software, such as firmware.

[0039] Although the present invention is described herein in conjunction with various embodiments, in the process of implementing the claimed invention, those skilled in the art may understand and implement other variations of the disclosed embodiments by viewing the drawings, the disclosure, and the appended claims. In the claims, the word "comprising" does not exclude other components or steps, and "one" or "an" does not exclude multiple situations. A single processor or other unit may implement several functions listed in the claims. Certain measures are recorded in different dependent claims, but this does not mean that these measures cannot be combined to produce good results.

[0040] The embodiments of the present application have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes are obvious to those of ordinary skill in the art without departing from the scope of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A fishing effort estimation method, characterized in that: include, Obtain the position data of multiple fishing vessels during the sample period; Preprocessing and extracting feature vectors of all the ship position data to obtain feature vectors of all the ship position data as ship position feature vectors; Constructing an SMA-lightGBM model, and using the ship position feature vector for training to obtain a fishing behavior recognition model; Obtain the position data of fishing vessels in the target sea area during the target period; The position data of fishing boats in the target sea area during the target period are pre-processed and feature vectors are extracted, and then input into the fishing behavior recognition model to identify the fishing behavior and time distribution of fishing boats in each fishing area of ​​the target sea area during the target period; The fishing behavior and time distribution of fishing boats in each fishing area in the target sea area during the target period are accumulated to obtain the fishing effort of each fishing area in the target sea area during the target period.

2. The method according to claim 1, characterized in that The step of constructing the SMA-lightGBM model and using the ship position feature vector for training to obtain a fishing behavior recognition model includes: Dividing all of the ship position feature vectors into a training set and a test set; The training set is randomly and evenly divided into a plurality of subsets; Using one of the subsets as a validation set and the remaining subsets as training sets; Each of the subsets is used as a validation set in turn, and the remaining subsets are used as training sets, and the SMA-lightGBM model is trained and verified to obtain an SMA-lightGBM model with optimal parameters as a fishing behavior recognition model; The fishing behavior recognition model is tested using the ship position feature vector in the test set to obtain average accuracy, sensitivity, specificity and Matthews correlation coefficient as generalization ability evaluation indicators of the fishing behavior recognition model.

3. The method according to claim 2, characterized in that The step of taking each of the subsets as a validation set in turn and the remaining subsets as training sets, training and validating the SMA-lightGBM model to obtain an SMA-lightGBM model with optimal parameters includes: Each of the subsets is used as a validation set, and the remaining subsets are used as training sets, to obtain multiple training set combinations; For each training set combination, the SMA parameters are continuously adjusted to train and verify the SMA-lightGBM model to obtain the average accuracy; The SMA-lightGBM model with the highest average accuracy is used as the SMA-lightGBM model with optimal parameters.

4. The method according to claim 3, characterized in that The step of continuously adjusting the SMA parameters for each training set combination to train and verify the SMA-lightGBM model to obtain an average accuracy rate includes: For each training set combination, perform the following steps respectively: Initialize SMA parameters, The ship position feature vector in the training set is input into the SMA-lightGBM model for training, The trained lightGBM model is verified using the ship position feature vector in the verification set to obtain the accuracy and change trend of the SMA-lightGBM model after SMA parameter optimization. Continue to adjust the population size (popsize) and maximum number of iterations (Max Miter) in the SMA parameters until the SMA-lightGBM model reaches the target accuracy. The average accuracy of the SMA-lightGBM model to achieve the target accuracy after adjusting the SMA parameters each time is calculated.

5. The method according to claim 1, characterized in that The step of preprocessing the position data of fishing boats in the target sea area during the target time period and extracting the feature vectors and then inputting them into the fishing behavior recognition model to identify the fishing behavior and time period distribution of fishing boats in each fishing area of ​​the target sea area during the target time period includes: Dividing the target time period into a plurality of unit time periods of equal duration; The position data of the fishing boats in the target sea area in each unit time period are pre-processed and feature vectors are extracted, and then input into the fishing behavior recognition model to identify the fishing behavior of the fishing boats in the target sea area in each unit time period; For each unit time period, the sea area distribution of each fishing area in the target sea area within the unit time period is obtained according to the position data and fishing behavior of fishing vessels in the target sea area; The fishing behaviors and time period distribution of fishing boats in the sea area distribution range of each fishing area in each of the unit time periods are summarized to obtain the fishing behaviors and time period distribution of fishing boats in each fishing area of ​​the target sea area in the target time period.

6. The method according to claim 5, characterized in that The step of obtaining the sea area distribution of each fishing area in the target sea area in each unit time period according to the position data and fishing behavior of fishing vessels in the target sea area, include, For each of the unit time periods, the following steps are performed respectively: The location of fishing boats in the target sea area is extracted based on their position data and fishing behavior. Select several positions of the fishing boat in the fishing state as reference positions. Calculate the distance between each fishing vessel other than the reference position and each of the reference positions, According to the distances between each of the reference positions and the fishing boats other than the reference position, the fishing boats having position correlation with each reference position are obtained. For each reference position, the position distribution range of the fishing boats having position correlation with each reference position is respectively used as the fishing area. The sea area distribution of each fishing area in the target sea area within the unit period is summarized.

7. The method according to claim 6, characterized in that The step of obtaining fishing boats having position correlation with each reference position based on the distances between each reference position and the fishing boats other than the reference position, include, The position of each fishing boat is classified into the same position group as the reference position with the shortest distance, Obtaining the geographic center of the distribution range of the positions of all fishing vessels in each of the position groups, The fishing boat position closest to the geographic center is used as the reference position after iteration. Determine whether the reference position after iteration has changed, If yes, then continue to calculate the geographic center and iteratively generate the reference position, If not, then several position groups are obtained. The reference position and the fishing boat whose positions are in the same position group are taken as the reference position and the fishing boat with position correlation.

8. The method according to claim 1, 5, 6 or 7, characterized in that: The step of accumulating the fishing behavior and time distribution of fishing boats in each fishing area of ​​the target sea area during the target time period to obtain the fishing effort of each fishing area in the target sea area during the target time period, include, For each fishing area, According to the fishing behavior and time distribution of fishing boats in the target period, the proportion of time each fishing boat is in the fishing state in the target period is obtained. The fishing effort of the fishing area during the target period is obtained by accumulating the proportion of time that all fishing vessels in the fishing area are in the fishing state during the target period; The fishing effort of each fishing area in the target sea area during the target period is summarized.

9. A method for detecting the distribution of fishery resources, characterized in that: include, Obtaining the fishing effort of each fishing area in the target sea area within the target period in the fishing effort estimation method according to any one of claims 1 to 8; The ratio of the fishing efforts in each fishing area of ​​the target sea area during the target period is taken as the distribution ratio of fishery resources in each fishing area of ​​the target sea area.

10. A fishery resource distribution detection system, characterized in that: include, A receiving unit, used for receiving the position data of multiple fishing vessels within a sample period; A fishing calculation unit, used to obtain the position data of multiple fishing vessels during the sample period; Preprocessing and extracting feature vectors of all the ship position data to obtain feature vectors of all the ship position data as ship position feature vectors; Constructing an SMA-lightGBM model, and using the ship position feature vector for training to obtain a fishing behavior recognition model; Obtain the position data of fishing vessels in the target sea area during the target period; The position data of fishing boats in the target sea area during the target period are pre-processed and feature vectors are extracted, and then input into the fishing behavior recognition model to identify the fishing behavior and time distribution of fishing boats in each fishing area of ​​the target sea area during the target period; The fishing behavior and time distribution of fishing boats in each fishing area of ​​the target sea area during the target period are accumulated to obtain the fishing effort of each fishing area in the target sea area during the target period; Fishery resource detection unit, used to obtain the fishing effort of each fishing area in the target sea area during the target period; The ratio of the fishing efforts in each fishing area of ​​the target sea area during the target period is taken as the distribution ratio of fishery resources in each fishing area of ​​the target sea area.

Images