A method and system for identifying fishing vessel operation modes based on neural networks.

Abstract

A method and system for identifying fishing vessel operation modes based on neural networks. The method includes: acquiring historical operation data of fishing vessels of different operation types; preprocessing the historical operation data to obtain feature data for each fishing vessel's operation; wherein the feature data includes speed data and latitude and longitude data; performing feature filtering and extraction on the feature data to construct an input feature vector; building a neural network model based on the input feature vector and the fishing vessel's operation mode; training the neural network model to obtain a fishing vessel operation mode identification model; wherein the input feature vector includes a speed ratio feature vector and a latitude and longitude feature vector; inputting the fishing vessel operation data to be identified into the fishing vessel operation mode identification model to obtain a fishing vessel operation mode identification result; if the identification result differs from the fishing vessel's registered operation mode, it is determined to be an illegal operation mode.

Description

Technical Field

[0001] This invention relates to the field of fishing vessel operation mode recognition technology, and in particular to a method and system for fishing vessel operation mode recognition based on neural networks. Background Technology

[0002] Currently, due to overfishing exceeding resource regeneration capacity and chaotic fishing operations, my country's marine fishery ecosystem is severely damaged. The legal and compliant operation methods and production order of fishing vessels directly affect the sustainable and cyclical development of resources in my country's marine ecosystem. At present, fishery administration and other relevant supervisory and management departments across my country mainly supervise fishing vessel operations through manual patrols, public reports, and random sampling inspections. Due to the wide distribution of my country's marine fishing grounds, the diversity of fishing vessel operations, and limited regulatory manpower, the supervision of fishing vessel operations often requires a huge investment of manpower, material resources, and financial resources, while the results are disproportionate to the investment and the efficiency is low. Summary of the Invention

[0003] Based on the above analysis, the present invention aims to provide a method and system for identifying fishing vessel operation modes based on neural networks, in order to solve the problem of low efficiency in existing methods for identifying fishing vessel operation modes.

[0004] On one hand, embodiments of the present invention provide a method for identifying fishing vessel operation modes based on neural networks, comprising the following steps:

[0005] Historical operation data of fishing vessels of different operation types are acquired, and the historical operation data is preprocessed to obtain characteristic data of each fishing vessel's operation; wherein, the characteristic data includes speed data and latitude and longitude data;

[0006] The feature data is filtered and extracted to construct an input feature vector; a neural network model is built based on the input feature vector and the fishing vessel operation mode, and the neural network model is trained to obtain a fishing vessel operation mode recognition model; the input feature vector includes a speed ratio feature vector and a latitude and longitude feature vector;

[0007] The fishing vessel operation data to be identified is input into the fishing vessel operation mode identification model to obtain the fishing vessel operation mode identification result. If the identification result is different from the fishing vessel's registered operation mode, it is determined to be an illegal operation mode.

[0008] The beneficial effects of the above technical solution are as follows:

[0009] By constructing and training a fishing vessel operation mode recognition model using historical fishing vessel operation data, the model can quickly and accurately identify the type of fishing vessel operation based on the trained data, achieving high recognition efficiency. Combining speed ratio features with latitude and longitude features can significantly improve recognition accuracy.

[0010] Based on the fundamental improvements to the above scheme, feature filtering and extraction are performed on the feature data to construct an input feature vector, including:

[0011] Calculate the percentage of speed in different speed ranges based on the speed data, and construct a speed percentage feature vector;

[0012] A fixed number of random samples are taken from the latitude and longitude data, and the main features of the latitude and longitude data are extracted using principal component analysis to obtain latitude and longitude feature vectors.

[0013] The beneficial effects of the above technical solution are as follows:

[0014] By sampling latitude and longitude data with a fixed number of random samples, the data characteristics are preserved while the data size is reduced. By using principal component analysis, the main features in the latitude and longitude features are extracted, thereby reducing the amount of data processing and improving efficiency without losing the main features and maintaining data accuracy.

[0015] Furthermore, the input feature vector also includes a speed change rate feature vector;

[0016] The process of filtering and extracting features from the feature data to construct an input feature vector also includes:

[0017] Calculate the average rate of change of speed in different speed ranges based on the speed data, and construct a feature vector of the rate of change of speed.

[0018] The beneficial effects of the above technical solution are as follows:

[0019] The accuracy of identifying fishing vessel operation methods can be further improved by adding a feature vector of the rate of change of speed.

[0020] Furthermore, the input feature vector also includes a heading change rate feature vector;

[0021] The process of filtering and extracting features from the feature data to construct an input feature vector also includes:

[0022] Calculate the rate of change of course based on the heading data; calculate the percentage of the rate of change of course in different heading change intervals, and construct a feature vector of the rate of change of course.

[0023] The beneficial effects of the above technical solution are as follows:

[0024] The accuracy of identifying fishing vessel operation methods can be further improved by adding a feature vector of the rate of change of heading.

[0025] Furthermore, the neural network model is a wavelet neural network model, which includes at least one hidden layer, and the activation function of the hidden layer is the POLYWOG wavelet function.

[0026] The beneficial effects of the above technical solution are as follows:

[0027] By employing a wavelet neural network model, the network structure is simpler, the convergence speed is faster, and the accuracy is higher.

[0028] On the other hand, embodiments of the present invention provide a fishing vessel operation mode identification system based on a neural network, comprising:

[0029] The data acquisition module is used to acquire historical operation data of fishing vessels of different operation types, preprocess the historical operation data, and obtain characteristic data of each fishing vessel's operation; wherein, the characteristic data includes speed data and latitude and longitude data;

[0030] The model training module is used to filter and extract features from the feature data to construct an input feature vector; build a neural network model based on the input feature vector and the fishing vessel operation mode, and train the neural network model to obtain a fishing vessel operation mode recognition model; the input feature vector includes a speed ratio feature vector and a latitude and longitude feature vector;

[0031] The fishing vessel operation mode identification module inputs the fishing vessel operation data to be identified into the fishing vessel operation mode identification model to obtain the fishing vessel operation mode identification result. If the identification result is different from the fishing vessel's registered operation mode, it is determined to be an illegal operation mode.

[0032] The beneficial effects of the above technical solution are as follows:

[0033] The data acquisition module uses historical fishing vessel operation data, while the model training module constructs and trains a fishing vessel operation mode recognition model. Based on the trained model, the operation mode recognition module can quickly and accurately identify the type of fishing vessel operation, achieving high recognition efficiency. By combining speed ratio features with latitude and longitude features, the recognition accuracy can be greatly improved.

[0034] Based on further improvements to the above technical solution, the model training module includes a feature extraction module, which includes:

[0035] The speed proportion feature extraction module is used to calculate the speed proportion of different speed ranges based on the speed data and construct a speed proportion feature vector.

[0036] The latitude and longitude feature extraction module is used to perform a fixed number of random samples on the latitude and longitude data, and to extract the main features of the latitude and longitude data using principal component analysis to obtain latitude and longitude feature vectors.

[0037] The beneficial effects of the above technical solution are as follows:

[0038] The feature extraction module samples latitude and longitude data using a fixed number of random samples, reducing the data size while preserving the data characteristics. By employing principal component analysis, it extracts the main features from the latitude and longitude features, thereby reducing the amount of data processing and improving efficiency without losing the main features and maintaining data accuracy.

[0039] Furthermore, the input feature vector also includes a speed change rate feature vector;

[0040] The feature extraction module further includes: a speed change rate feature extraction module, used to calculate the average speed change rate in different speed ranges based on the speed data, and construct a speed change rate feature vector.

[0041] The beneficial effects of the above technical solution are as follows: by adding the feature vector of the rate of change of speed, the accuracy of identifying the fishing vessel operation mode is further improved.

[0042] Furthermore, the input feature vector also includes a heading change rate feature vector;

[0043] The feature extraction module further includes: a heading change rate feature extraction module, used to calculate the heading change rate based on heading data; calculate the proportion of heading change rate in different heading change intervals, and construct a heading change rate feature vector.

[0044] The beneficial effects of the above technical solution are as follows: by adding the heading change rate feature vector, the accuracy of identifying fishing vessel operation methods is further improved.

[0045] Furthermore, the neural network model is a wavelet neural network model, which includes at least one hidden layer, and the activation function of the hidden layer is the POLYWOG wavelet function.

[0046] The beneficial effects of the above technical solution are as follows: The model training module adopts a wavelet neural network model, which has a simpler network structure, faster convergence speed, and higher accuracy.

[0047] In this invention, the above-described technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of this invention will be set forth in the following description, and some advantages may become apparent from the description or be learned by practicing the invention. The objects and other advantages of this invention can be realized and obtained from what is particularly pointed out in the description and drawings. Attached Figure Description

[0048] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts.

[0049] Figure 1This is a flowchart of a neural network-based method for identifying fishing vessel operation modes according to an embodiment of the present invention;

[0050] Figure 2 A chart showing the percentage of speeds for fishing vessels of different types;

[0051] Figure 3 Latitude and longitude trajectory maps for fishing vessels of different types;

[0052] Figure 4 This is a structural block diagram of the fishing vessel operation mode recognition system based on neural networks, according to an embodiment of the present invention. Detailed Implementation

[0053] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0054] A specific embodiment of the present invention discloses a method for identifying fishing vessel operation modes based on neural networks, such as... Figure 1 This includes the following steps:

[0055] S1. Obtain historical operation data of fishing vessels of different operation types, preprocess the historical operation data to obtain characteristic data of each fishing vessel's operation; wherein, the characteristic data includes speed data and latitude and longitude data.

[0056] According to China's "Regulations on the Administration of Fishery Fishing Licenses," fishing gear in my country is mainly divided into nine categories: trawlers, gillnets, sea nets, traps, angling tackle, miscellaneous fishing gear, purse seines, shovels, and traps. Because the installation information of onboard positioning clients for fishing vessels using shovels and traps is generally lacking compared to other fishing methods, a specific embodiment of this invention includes historical fishing data for seven categories of fishing vessels: trawlers, gillnets, traps, angling tackle, miscellaneous fishing gear, and purse seines.

[0057] During implementation, historical operational data of fishing vessels can be obtained from the Automatic Identification System (AIS) or the fishing vessel monitoring system.

[0058] The sample size of the seven types of fishing vessels is shown in Table 1.

[0059] Table 1

[0060]

[0061]

[0062] Historical data includes flight speed data and latitude / longitude data.

[0063] Studies have shown that speed is highly sensitive to fishing methods, with significant differences in the proportion of speed used by different types of fishing vessels, while speed data for similar types of vessels exhibit similarities. For example... Figure 2 As shown in Figure 2, (a) to (g) represent the speed percentages of fishing vessels operating under seven different fishing methods: trawl, gillnet, seine, purse seine, miscellaneous fishing gear, angling, and traps / pots. The operating speeds of trawlers are mainly concentrated in the range of (0.5, 4.5) knots; those of gillnetters are mainly concentrated in the range of (0, 0.5) knots; those of gillnetters are mainly concentrated in the range of (0.5, 4.5) knots, with a more pronounced speed range of (5.5, 12) knots compared to purse seines and trawlers; the operating speeds of purse seines are mainly concentrated in the range of (0.5, 4.5) knots; the operating speeds of miscellaneous fishing gear vessels are mainly concentrated in the ranges of (0.5, 4.5) knots and (5.5, 12) knots; the operating speeds of angling tackle vessels are mainly concentrated in the ranges of (0, 0.5) knots and (5.5, 12) knots; and the operating speeds of angling tackle vessels show a relatively significant proportion within the ranges of (0, 2.5) knots, (2.5, 5.5) knots, and (5.5, 12) knots.

[0064] Research has revealed significant differences in the trajectories of fishing vessels operating under different conditions. For example... Figure 3 As shown, (a), (b), and (c) are the trajectories and trajectory feature maps of trawlers, gillnets, and nets, respectively. It can be seen that the trajectory map of trawlers forms multiple concentrated areas, while the trajectory map of gillnets, with its clearly defined target location, is mostly a straight line. Its latitude and longitude characteristics are relatively stable within a certain area and exhibit an approximately periodic nature. The speed ratio of gillnet fishing vessels is similar to that of trawlers, and their course trajectories are also very similar. Therefore, combining the speed ratio feature with latitude and longitude features can effectively distinguish fishing vessels of different operation types, and the computational efficiency is high.

[0065] Due to the instability of the operating environment and the diversity of the marine environment, the speed data of various types of fishing vessels fluctuates greatly, but they generally have a stable operating speed. In this embodiment, a maximum speed v is set. max =25 sections, thereby eliminating some numerical outliers. At the same time, duplicate and abnormal data in the original data are removed by designing deduplication functions and noise reduction rules to prevent them from affecting subsequent feature extraction.

[0066] The original latitude and longitude data are coordinate points (x, y), where x represents longitude and y represents latitude, such as (122.2345, 33.3241). During data preprocessing, the 2D latitude and longitude data is mapped to 1D feature data through principal component analysis, which facilitates the extraction of latitude and longitude feature vectors.

[0067] S2. Perform feature filtering and extraction on the feature data to construct an input feature vector; build a neural network model based on the input feature vector and the fishing vessel operation mode label, and train the neural network model to obtain a fishing vessel operation mode recognition model; the input feature vector includes a speed ratio feature vector and a latitude and longitude feature vector.

[0068] Specifically, feature selection and extraction includes the following two steps:

[0069] S21. Calculate the percentage of speed in different speed ranges based on the speed data, and construct a speed percentage feature vector.

[0070] Specifically, based on the accuracy of the commonly available speed data, the speed range of 0-25 knots is divided into 250 intervals with a 0.1 knot increment. The percentage of speed in each interval is calculated, and a speed percentage feature vector is constructed. During implementation, analysis of fishing vessel operation data revealed that the amount of data for speeds greater than 10 knots is extremely limited. Within the 10-25 knot speed range, there is no significant difference in the speed percentage data for different types of fishing vessels. Therefore, to maintain data characteristics and minimize computational load, the first 100 dimensions of the speed percentage feature are extracted as the primary features and normalized to form the speed percentage feature vector.

[0071] S22. A fixed number of random samples are taken from the latitude and longitude data, and the main features of the latitude and longitude data are extracted using principal component analysis to obtain latitude and longitude feature vectors.

[0072] Due to the large volume of latitude and longitude data, in order to maintain data characteristics and reduce computational scale, a fixed number of random samples were first taken from the preprocessed latitude and longitude data to ensure that the sample size of each sample was consistent. Principal component analysis was then used to extract the main features of the latitude and longitude data from the sampled data, and the resulting feature vectors were obtained after normalization.

[0073] The input feature vector is composed of the speed ratio feature vector and the latitude and longitude feature vector. A neural network model is built based on the dimension of the input feature vector and the fishing vessel operation mode label. The input of the model is the input feature vector, and the output is the fishing vessel operation mode. The mini-batch stochastic gradient descent algorithm is used to train the network model.

[0074] Because wavelet function families possess compact support and near-orthogonality, there is no need to add a Dropout layer in the neural network model to prevent overfitting. The model maintains its generalization ability while effectively improving training speed. The neural network model can be a wavelet neural network model, which includes at least one hidden layer. The hidden layer activation function uses the POLYWOG wavelet function. In the wavelet neural network, the activation function of each neuron is different, and the output y of the k-th neuron... k It can be represented as:

[0075]

[0076] Among them, a k b is the scaling factor of the activation wavelet function of the k-th neuron. k S is the offset of the activation wavelet function of the k-th neuron. k It is the weighted sum of the input features of the k-th neuron, and ψ(S) is the POLYWOG wavelet function.

[0077] The model was trained and tested using operational data from 951 fishing vessels listed in Table 1. Five different sets of sample data were randomly collected for testing, and the classification results are shown in Table 2. Comparing the results of the five test sets, it can be seen that the recognition rates for all types of fishing vessels are high and stable. The overall recognition accuracy for each test set was calculated, and the final model achieved an average recognition accuracy of 99.26% across the five test sets. By combining speed ratio features with latitude and longitude features, the operational methods of fishing vessels can be accurately identified.

[0078] Table 2 Classification results of neural network models

[0079]

[0080] Analysis of data from different types of fishing vessels revealed significant differences in their speed variation rates. Trawlers exhibited relatively low speed variations during operation, primarily operating at low speeds. Set net vessels showed significantly higher speed variations than trawlers, with more frequent high-low speed changes. Gillnet vessels also primarily operated at low speeds, but their speed variations were greater than those of trawlers. Purse seine vessels operated at very low speeds with minimal speed changes, but significant speed variations occurred at the beginning and end of the operation. Fishing tackle vessels displayed a periodic speed variation trend. Cage and pot vessels primarily operated at low to medium speeds, with more pronounced speed variations compared to purse seine vessels. Miscellaneous fishing tackle vessels exhibited relatively stable speed variations during operation.

[0081] Therefore, the rate of change of airspeed can be used as a feature for classification and recognition, thereby further improving the accuracy of classification and recognition.

[0082] Therefore, the input feature vector also includes a speed change rate feature vector; that is, as an optional implementation, the constructed input feature vector includes a speed ratio feature vector, a latitude and longitude feature vector, and a speed change rate feature vector, and the neural network model is trained in combination with the fishing vessel operation mode to obtain a fishing vessel operation mode recognition model.

[0083] Calculate the average rate of change of speed in different speed ranges based on the speed data, and construct a feature vector of the rate of change of speed.

[0084] Specifically, the rate of change of speed is calculated based on the speed data. The rate of change of speed = the change in speed (increase or decrease) divided by the original speed value.

[0085] Similar to step S21, the speed range of 0-25 knots is divided into 250 intervals with 0.1 knots as the increments. The average speed change rate in each speed interval is calculated, and a speed change rate feature vector is constructed. During implementation, analysis of fishing vessel operation data revealed that the amount of data for fishing vessel speeds greater than 10 knots is extremely limited. Furthermore, within the 10-25 knot speed range, there is no significant difference in the speed change rate data among different types of fishing vessels. Therefore, to maintain data characteristics and minimize computational load, the first 100 dimensions of the speed change rate feature are extracted as the primary features and normalized to form the speed change rate feature vector.

[0086] Research and analysis of data from different types of fishing vessels revealed significant differences in course changes. For example, trawlers and gillnetters exhibit frequent course changes, with a majority of course change rates ranging from -0.16 to 0.16. Gillnetters show very little course change, with course change rates mainly concentrated around 0 and 0.5. Purse seine vessels maintain an almost straight course, with course change rates concentrated around 0. Using the course change rate as a classification feature further improves the accuracy of classification.

[0087] Therefore, the input feature vector also includes a heading change rate feature vector; that is, as an optional implementation, the constructed input feature vector includes a speed ratio feature vector, a latitude and longitude feature vector, and a heading change rate feature vector; the neural network model is trained in combination with the fishing vessel operation mode to obtain a fishing vessel operation mode recognition model.

[0088] As an optional implementation, the constructed input feature vector includes a speed percentage feature vector, a latitude and longitude feature vector, a heading change rate feature vector, and a speed change rate feature vector; the neural network model is trained in conjunction with the fishing vessel operation mode to obtain a fishing vessel operation mode recognition model.

[0089] The eigenvector of the rate of change of heading is extracted through the following steps:

[0090] The rate of change of heading is calculated based on the heading data. The rate of change of heading equals the changed heading value divided by 360 degrees. The percentage of the rate of change of heading in different heading change intervals is calculated, and a feature vector of the rate of change of heading is constructed. In practice, the minimum value of the heading change is used as the calculation value; for example, +270° = -90°, and 90 is taken as the change value. Therefore, the numerical range of the rate of change of heading is between -0.5 and 0.5. The speed change rate interval is divided into 10 features with intervals of 0.1, and the value corresponding to each feature is the percentage of the rate of change of heading in that interval.

[0091] S3. Input the fishing vessel operation data to be identified into the fishing vessel operation mode identification model to obtain the fishing vessel operation mode identification result. If the identification result is different from the operation mode registered by the fishing vessel, it is determined to be an illegal operation mode.

[0092] In practice, the speed and latitude / longitude data of the fishing vessel to be identified can be extracted. Following the method used in steps S1 and S2 to construct the input feature vector, the input feature vector of the fishing vessel to be identified can be extracted and input into the trained fishing vessel operation mode recognition model to obtain the recognition result. To improve recognition accuracy, the speed change rate feature vector and / or heading change rate feature vector of the fishing vessel to be identified can be further extracted as part of the input feature vector to obtain the recognition result.

[0093] During implementation, the intelligent identification results of fishing vessel data by the model can be displayed on the terminal, assisting fishery administration and other relevant departments in supervising and enforcing regulations on the production order of fishing vessel operations in my country's oceans. For example, the fishing vessel operation methods identified by the model in step S3 can be visualized on the terminal using frameworks and technologies such as Vue and JavaScript. Fishing vessels identified by the model as engaging in illegal operations are highlighted, and a comparison between the illegal operation methods and the registered operation methods is provided, offering auxiliary suggestions for fishery administration and other relevant regulatory departments in various regions to supervise and enforce regulations on the production order of fishing vessel operations in my country's oceans. The neural network-based fishing vessel operation method identification method provided by the embodiments of this invention is simple and easy to implement, greatly reducing the human, material, and financial resources required for the supervision of fishing vessel operation methods, and is accurate and efficient.

[0094] A specific embodiment of the present invention discloses a fishing vessel operation mode recognition system based on neural networks, such as... Figure 4 As shown, it includes:

[0095] The data acquisition module is used to acquire historical operation data of fishing vessels of different operation types, preprocess the historical operation data to obtain characteristic data of each fishing vessel's operation; wherein, the characteristic data includes speed data and latitude and longitude data. Specific data acquisition and processing methods are described in the foregoing embodiments and will not be repeated here.

[0096] The model training module is used to filter and extract features from the feature data to construct an input feature vector; based on the input feature vector and the fishing vessel's registered operation mode, a neural network model is built, and the neural network model is trained to obtain a fishing vessel operation mode recognition model; the input feature vector includes a speed ratio feature vector and a latitude and longitude feature vector. Specific feature extraction and model training methods are described in the aforementioned embodiments and will not be repeated here.

[0097] The fishing vessel operation mode identification module inputs the fishing vessel operation data to be identified into the fishing vessel operation mode identification model to obtain the fishing vessel operation mode identification result. If the identification result is different from the fishing vessel's registered operation mode, it is determined to be an illegal operation mode. For the specific identification process and method, please refer to the aforementioned embodiments, which will not be repeated here.

[0098] Preferably, the model training module includes a feature extraction module, which includes:

[0099] The speed percentage feature extraction module is used to calculate the speed percentage of different speed ranges based on the speed data and construct a speed percentage feature vector. The latitude and longitude feature extraction module is used to perform a fixed number of random samples on the latitude and longitude data and use principal component analysis to extract the main features of the latitude and longitude data to obtain a latitude and longitude feature vector. The specific feature extraction process is described in the aforementioned embodiment and will not be repeated here.

[0100] Preferably, the input feature vector further includes a speed change rate feature vector; the feature extraction module further includes a speed change rate feature extraction module, used to calculate the average speed change rate in different speed ranges based on the speed data, and construct a speed change rate feature vector. The specific speed change rate feature extraction process is described in the foregoing embodiments and will not be repeated here.

[0101] Preferably, the input feature vector further includes a heading change rate feature vector; the feature extraction module further includes: a heading change rate feature extraction module, used to calculate the heading change rate based on the heading data; calculate the proportion of the heading change rate in different heading change intervals, and construct a heading change rate feature vector. The specific heading change rate feature extraction process is described in the foregoing embodiments and will not be repeated here.

[0102] Preferably, the neural network model is a wavelet neural network model, which includes at least one hidden layer, and the activation function of the hidden layer is the POLYWOG wavelet function.

[0103] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0104] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A neural network-based fishing boat operation mode recognition method, characterized by, Includes the following steps: Historical fishing data of fishing vessels of different types are acquired, and the historical fishing data is preprocessed to obtain characteristic data of each fishing vessel's operation; wherein, the characteristic data includes speed data and latitude and longitude data; the different types of fishing vessels include trawlers, gillnetters, sea netters, fish traps, fishing tackle, miscellaneous fishing tackle, and purse seine fishing vessels; The feature data is filtered and extracted to construct an input feature vector; a neural network model is built based on the input feature vector and the fishing vessel operation mode, and the neural network model is trained to obtain a fishing vessel operation mode recognition model; the input feature vector includes a speed ratio feature vector and a latitude and longitude feature vector; The fishing vessel operation data to be identified is input into the fishing vessel operation mode identification model to obtain the fishing vessel operation mode identification result. If the identification result is different from the operation mode registered by the fishing vessel, it is determined to be an illegal operation mode. The feature data is subjected to feature filtering and extraction to construct an input feature vector, including: Calculate the percentage of speed in different speed ranges based on the speed data, and construct a speed percentage feature vector; A fixed number of random samples are taken from the latitude and longitude data, and the main features of the latitude and longitude data are extracted using principal component analysis to obtain latitude and longitude feature vectors. The input feature vector also includes a speed change rate feature vector; The process of filtering and extracting features from the feature data to construct an input feature vector also includes: calculating the average rate of change of speed in different speed ranges based on the speed data, and constructing a rate of change of speed feature vector. The input feature vector also includes a heading change rate feature vector; The process of filtering and extracting features from the feature data to construct an input feature vector also includes: Calculate the rate of change of course based on the heading data; calculate the percentage of the rate of change of course in different heading change intervals, and construct a feature vector of the rate of change of course; The speed range of 0-25 is divided into 250 intervals with an interval of 0.1 knots. The speed ratio in each speed range is calculated, and the first 100 dimensions of the speed ratio feature are extracted as the main features. After normalization, they are used as the speed ratio feature vector. The speed range of 0 to 25 knots was divided into 250 intervals with an interval of 0.1 knots. The average speed change rate in each speed range was calculated. The first 100 dimensions of the speed change rate feature were extracted as the main features and normalized to form the speed change rate feature vector. The rate of change of heading is equal to the change in heading value divided by 360 degrees. The rate of change of heading is divided into 10 features at intervals of 0.

1. The value corresponding to each feature is the proportion of the rate of change of heading in that interval, forming a feature vector of the rate of change of heading.

2. The method for identifying fishing vessel operation modes based on neural networks according to claim 1, characterized in that, The neural network model is a wavelet neural network model, which includes at least one hidden layer, and the activation function of the hidden layer is the POLYWOG wavelet function.

3. A fishing vessel operation mode recognition system based on neural networks, characterized in that, include: The data acquisition module is used to acquire historical operation data of fishing vessels of different operation types, preprocess the historical operation data to obtain characteristic data of each fishing vessel's operation; wherein, the characteristic data includes speed data and latitude and longitude data; the different operation types of fishing vessels include trawlers, gillnetters, sea netters, fish traps, fishing tackle, miscellaneous fishing tackle and purse seine fishing vessels; The model training module is used to filter and extract features from the feature data to construct an input feature vector; build a neural network model based on the input feature vector and the fishing vessel operation mode, and train the neural network model to obtain a fishing vessel operation mode recognition model; the input feature vector includes a speed ratio feature vector and a latitude and longitude feature vector; The fishing vessel operation mode identification module inputs the fishing vessel operation data to be identified into the fishing vessel operation mode identification model to obtain the fishing vessel operation mode identification result. If the identification result is different from the fishing vessel's registered operation mode, it is determined to be an illegal operation mode. The model training module includes a feature extraction module, which includes: The speed proportion feature extraction module is used to calculate the speed proportion of different speed ranges based on the speed data and construct a speed proportion feature vector. The latitude and longitude feature extraction module is used to perform a fixed number of random samples on the latitude and longitude data, and to extract the main features of the latitude and longitude data using principal component analysis to obtain latitude and longitude feature vectors. The input feature vector also includes a speed change rate feature vector; The feature extraction module further includes: a speed change rate feature extraction module, used to calculate the average speed change rate in different speed ranges based on the speed data, and construct a speed change rate feature vector. The input feature vector also includes a heading change rate feature vector; The feature extraction module further includes: a heading change rate feature extraction module, used to calculate the heading change rate based on the heading data; calculate the proportion of heading change rate in different heading change intervals; and construct a heading change rate feature vector. The speed range of 0-25 is divided into 250 intervals with an interval of 0.1 knots. The speed ratio in each speed range is calculated, and the first 100 dimensions of the speed ratio feature are extracted as the main features. After normalization, they are used as the speed ratio feature vector. The speed range of 0 to 25 knots was divided into 250 intervals with an interval of 0.1 knots. The average speed change rate in each speed range was calculated. The first 100 dimensions of the speed change rate feature were extracted as the main features and normalized to form the speed change rate feature vector. The rate of change of heading is equal to the change in heading value divided by 360 degrees. The rate of change of heading is divided into 10 features at intervals of 0.

1. The value corresponding to each feature is the proportion of the rate of change of heading in that interval, forming a feature vector of the rate of change of heading.

4. The fishing vessel operation mode recognition system based on neural networks according to claim 3, characterized in that, The neural network model is a wavelet neural network model, which includes at least one hidden layer, and the activation function of the hidden layer is the POLYWOG wavelet function.

Images