Job fishing boat navigation track prediction method, device, equipment, medium and product

Abstract

The application provides a kind of working fishing boat navigation track prediction method, device, equipment, medium and product, belong to trajectory prediction technical field, method includes: obtaining the AIS data of to-be-predicted working fishing boat;Input AIS data to pre-training completed trajectory prediction model, obtain the predicted track of to-be-predicted working fishing boat.Wherein, the training step of trajectory prediction model includes: obtaining the historical AIS data of working fishing boat, according to historical AIS data to obtain standard AIS data;Based on attention mechanism and multi-scale mechanism, improve CNN-LSTM model, obtain improved CNN-LSTM model;Based on standard AIS data, the improved CNN-LSTM model is trained, and the trajectory prediction model is obtained.The application improves CNN-LSTM model through attention mechanism and multi-scale mechanism, and trains the improved CNN-LSTM model to obtain the trajectory prediction model, which can improve the performance in the navigation track prediction of fishing boat, so that the prediction result is more accurate and reliable.

Description

Technical Field

[0001] This invention relates to the field of trajectory prediction technology, and in particular to a method, apparatus, equipment, medium and product for predicting the navigation trajectory of fishing vessels. Background Technology

[0002] Fishing vessel trajectory prediction is based on historical trajectory data. Through analysis and modeling, it predicts the movement trajectory of fishing vessels over a future period. This task is of great significance for improving vessel management, maritime safety, and the efficiency of fishery resource utilization. Traditional fishing vessel trajectory prediction methods are usually based on statistical or physical models, but these methods often rely on domain expert knowledge and artificial feature engineering, which have certain limitations when dealing with complex marine environments and variable fishery activities.

[0003] Currently, methods for predicting and analyzing fishing vessel trajectories mainly rely on deep learning feature extraction techniques. This method typically models the behavior of fishing vessels and extracts meaningful features from the raw trajectory data to support subsequent prediction tasks. However, this method does not fully consider the temporal relationships within the trajectory data, and the accuracy of the prediction model needs improvement. Summary of the Invention

[0004] This invention provides a method, apparatus, equipment, medium, and product for predicting the navigation trajectory of fishing vessels, in order to address the shortcomings of existing technologies that do not fully consider the temporal relationships in trajectory data and whose prediction model accuracy needs to be improved.

[0005] This invention provides a method for predicting the navigation trajectory of a fishing vessel, comprising the following steps: acquiring AIS data of the fishing vessel to be predicted; inputting the AIS data into a pre-trained trajectory prediction model to obtain the predicted trajectory of the fishing vessel to be predicted.

[0006] The training steps of the trajectory prediction model include: acquiring historical AIS data of the fishing vessel, obtaining standard AIS data based on the historical AIS data; improving the CNN-LSTM model based on attention mechanism and multi-scale mechanism to obtain an improved CNN-LSTM model; and training the improved CNN-LSTM model based on the standard AIS data to obtain the trajectory prediction model.

[0007] According to the present invention, a method for predicting the navigation trajectory of a fishing vessel is provided. The CNN-LSTM model includes a CNN module and an LSTM module connected to the output of the CNN module. The method for improving the CNN-LSTM model based on attention mechanism and multi-scale mechanism to obtain an improved CNN-LSTM model includes the following steps.

[0008] An attention module is set at the input of the CNN module. The attention module is used to determine the attention weight of the standard AIS data or the AIS data at each time step based on the content-based attention mechanism. The standard AIS data is then weighted and summed according to the attention weight to obtain the attention-adjusted standard AIS data.

[0009] By adjusting the convolution kernel parameters or pooling layer parameters of the CNN module, a multi-scale CNN module is obtained. The multi-scale CNN module is used to extract the spatial features of the attention-adjusted standard AIS data at different time scales and perform feature fusion to obtain multi-scale fused features.

[0010] Based on the attention module, the multi-scale CNN module, and the LSTM module, the improved CNN-LSTM model is obtained, and the LSTM module is used to extract the temporal features of the multi-scale fusion features.

[0011] According to the method for predicting the navigation trajectory of a fishing vessel provided by the present invention, the calculation formula for the weighted summation operation of the standard AIS data based on the attention weight is as follows:

[0012]

[0013] Where softmax is the activation function, W is the attention weight, feature_t is the standard AIS data at time step t, and N is the total number of time steps.

[0014] According to the present invention, a method for predicting the navigation trajectory of a fishing vessel is provided. The input data of the multi-scale CNN module consists of a three-dimensional tensor composed of data height, data width, and data channels. The multi-scale CNN module includes multiple convolutional layers, activation functions, pooling layers, and fully connected layers. The convolutional layers are used to perform convolution operations based on cross-correlation. The activation functions are used to perform nonlinear transformations on the feature maps output by the convolutional layers. The pooling layers are used to perform downsampling operations on the feature maps output by the convolutional layers based on max pooling.

[0015] According to a method for predicting the navigation trajectory of a fishing vessel provided by the present invention, the step of obtaining standard AIS data based on historical AIS data includes: cleaning the historical AIS data; formatting the cleaned historical AIS data based on a preset data format, wherein the preset data format includes a unique vessel identifier, voyage, longitude, latitude, heading, and upload time; and normalizing the formatted historical AIS data to obtain the standard AIS data.

[0016] The formula for normalization is as follows:

[0017]

[0018] Where x represents historical AIS data. σ is the mean of the historical AIS data, and σ is the standard deviation of the historical AIS data.

[0019] According to a method for predicting the navigation trajectory of a fishing vessel provided by the present invention, after obtaining the predicted trajectory of the fishing vessel to be predicted, the method further includes: acquiring the operational environment data and current operational target of the fishing vessel to be predicted; and performing trajectory anomaly detection on the fishing vessel to be predicted based on the operational environment data, the current operational target, and the predicted trajectory.

[0020] The present invention also provides a device for predicting the navigation trajectory of fishing vessels, including a training module, an acquisition module and a prediction module.

[0021] The training module is used to acquire historical AIS data of fishing vessels and obtain standard AIS data based on the historical AIS data; improve the CNN-LSTM model based on the attention mechanism and multi-scale mechanism to obtain an improved CNN-LSTM model; train the improved CNN-LSTM model based on the standard AIS data to obtain the trajectory prediction model.

[0022] The acquisition module is used to acquire AIS data of the fishing vessels to be predicted.

[0023] The prediction module is used to input the AIS data into a pre-trained trajectory prediction model to obtain the predicted trajectory of the fishing vessel to be predicted.

[0024] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the fishing vessel navigation trajectory prediction method as described above.

[0025] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the fishing vessel navigation trajectory prediction method as described above.

[0026] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the fishing vessel navigation trajectory prediction method as described above.

[0027] The present invention provides a method, apparatus, equipment, medium and product for predicting the navigation trajectory of fishing vessels. By improving the CNN-LSTM model through attention mechanism and multi-scale mechanism, the trajectory prediction model is obtained by training the improved CNN-LSTM model, which can improve the performance of fishing vessel navigation trajectory prediction and make the prediction results more accurate and reliable. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0029] Figure 1 This is a flowchart illustrating the method for predicting the navigation trajectory of fishing vessels provided by the present invention.

[0030] Figure 2 This is a schematic diagram of the structure of the fishing vessel navigation trajectory prediction device provided by the present invention.

[0031] Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0033] The terms "first," "second," etc., used in this invention are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class, without limiting the number of objects; for example, a first object can be one or more. Furthermore, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0034] All actions involving the acquisition of signals, information, or data in this invention are carried out in accordance with the relevant data protection laws and policies of the country where the device is located, and with the authorization granted by the owner of the relevant device.

[0035] The following is combined Figures 1-3 This invention describes the method, apparatus, equipment, medium, and product for predicting the navigation trajectory of fishing vessels.

[0036] Figure 1 This is a flowchart illustrating the method for predicting the navigation trajectory of fishing vessels provided by the present invention, as shown below. Figure 1 As shown, the present invention provides a method for predicting the navigation trajectory of fishing vessels. The present invention is applied to the real-time monitoring and early warning system of fishing vessels during fishing operations to improve the accuracy of prediction of the navigation trajectory of fishing vessels and the monitoring efficiency, including but not limited to the following steps S100-S200.

[0037] Step S100: Obtain Automatic Identification System (AIS) data of the fishing vessel to be predicted; AIS data is obtained by the fishing vessel transmitting data back at specified time intervals.

[0038] Step S200: Input the AIS data into the pre-trained trajectory prediction model to obtain the predicted trajectory of the fishing vessel to be predicted. Before inputting the AIS data into the trajectory prediction model, the AIS data is cleaned and formatted. Data cleaning includes removing outliers, filling in missing values, and smoothing noise to ensure data quality and consistency. Formatting refers to converting the cleaned AIS data into an input format acceptable to the trajectory prediction model, transforming the formatted AIS data into standardized data. The trajectory prediction model performs forward propagation to generate the predicted future trajectory. In the above process, the trajectory prediction model automatically extracts features from the input and performs pattern matching based on historical data and existing learning to generate the predicted output.

[0039] The training steps of the trajectory prediction model include steps S010-S030.

[0040] Step S010: Obtain historical AIS data of the fishing vessel and obtain standard AIS data based on the historical AIS data.

[0041] Step S020 involves improving the CNN-LSTM model based on attention and multi-scale mechanisms to obtain an improved CNN-LSTM model. Convolutional Neural Networks (CNNs) can extract spatial features and perform local and global feature learning on trajectory data, while Long Short-Term Memory (LSTM) networks can capture long-term dependencies in time-series data. By introducing attention mechanisms and multi-scale combinations, the performance of the CNN-LSTM model in predicting fishing vessel trajectories can be improved, making the prediction results more accurate and reliable. This improved CNN-LSTM method has better automated learning capabilities, learning more effective features and patterns from the marine environment and historical behavior of fishing vessels. Its predictive performance has been verified in experiments and has broad application prospects in fishing vessel management and maritime safety.

[0042] Specifically, all historical AIS data can be aggregated into a single dataset, or it can be divided into multiple subsets based on factors such as the fishing vessel's operational objectives and patterns. The improved CNN-LSTM model uses multiple network branches based on the number of subsets. After training the trajectory prediction model, the operational objectives, patterns, and AIS data of the fishing vessel to be predicted are acquired and input into the trajectory prediction model. Trajectory prediction is achieved by calling the corresponding network branches. The network structure of each branch is smaller than the overall trajectory prediction model, reducing computational power and improving prediction efficiency during application. During training, factors such as the fishing vessel's operational objectives and patterns can also be used to adjust attention weights.

[0043] Step S030: Train the improved CNN-LSTM model based on the standard AIS data to obtain the trajectory prediction model. Specifically, the training process may include selecting mean squared error (MSE) as the loss function, the Adam optimizer, and determining the number of training epochs to be 200 epochs with an initial learning rate of 0.01.

[0044] It is understood that this invention improves the CNN-LSTM model through attention and multi-scale mechanisms. The trajectory prediction model is obtained by training the improved CNN-LSTM model, which can improve the performance of fishing boat navigation trajectory prediction and make the prediction results more accurate and reliable.

[0045] Based on the above embodiments, as an optional embodiment, obtaining standard AIS data from the historical AIS data includes steps S011-S013.

[0046] Step S011: Perform data cleaning on the historical AIS data. Data cleaning includes removing outliers, filling in missing values, and smoothing noise to ensure data quality and consistency.

[0047] Step S012: The cleaned historical AIS data is formatted and converted based on a preset data format, which includes the ship's unique identifier, voyage, longitude, latitude, heading, and upload time.

[0048] Step S013: Normalize the formatted and converted historical AIS data to obtain the standard AIS data, which is the movement trajectory sequence of the fishing vessel.

[0049] In step S013, the normalization formula is as follows:

[0050]

[0051] Where x represents historical AIS data. σ is the mean of the historical AIS data, and σ is the standard deviation of the historical AIS data.

[0052] It is understood that by preprocessing historical AIS data, the present invention can effectively improve the accuracy of model training. By formatting and converting the cleaned historical AIS data based on a preset data format, data from different ships or different sources can be integrated into a unified format, thus expanding the applicability of the trajectory prediction model.

[0053] Based on the above embodiments, as an optional embodiment, the CNN-LSTM model includes a CNN module and an LSTM module connected to the output of the CNN module. The improvement of the CNN-LSTM model based on the attention mechanism and the multi-scale mechanism to obtain the improved CNN-LSTM model includes steps S021-S023.

[0054] Step S021: An attention module is set at the input of the CNN module. This attention module is used to determine the attention weight of the standard AIS data or the AIS data at each time step using a content-based attention mechanism. The standard AIS data is then weighted and summed according to the attention weights to obtain attention-adjusted standard AIS data. Specifically, the content-based attention mechanism can be a soft attention mechanism. In other embodiments, the weights of the time steps can also be adjusted based on factors such as the fishing vessel's operational objectives or operational mode.

[0055] Attention weights can be adaptively adjusted based on the standard AIS data or the AIS data to highlight important information and suppress noise or irrelevant information.

[0056] Step S022: Adjust the convolution kernel parameters or pooling layer parameters of the CNN module to obtain a multi-scale CNN module. This multi-scale CNN module is used to extract spatial features of the attention-adjusted standard AIS data at different time scales and perform feature fusion to obtain multi-scale fused features. Specifically, by adjusting the convolution kernel parameters or pooling layer parameters of the CNN module, features at different scales can be obtained.

[0057] Step S023: Based on the attention module, the multi-scale CNN module, and the LSTM module, the improved CNN-LSTM model is obtained, wherein the LSTM module is used to extract the temporal features of the multi-scale fusion features.

[0058] Understandably, this invention uses a CNN to extract spatial features from historical trajectory data of ships, then constructs an LSTM network to process time-series data, capturing temporal features and performing time-series modeling of the movement trends and historical behavior patterns of fishing vessels. The fishing vessel trajectory data is then input into the CNN-LSTM model for training. By introducing an attention mechanism into the CNN-LSTM model, the model can better utilize key information from historical trajectory data in the task of predicting fishing vessel trajectories, improving prediction performance and accuracy. Furthermore, introducing a multi-scale feature fusion module into the CNN-LSTM model allows it to better capture information at different time scales, improving its understanding and prediction capabilities of complex features in fishing vessel trajectory data, resulting in better performance and capabilities for the task of predicting fishing vessel trajectories.

[0059] Based on the above embodiments, as an optional embodiment, the calculation formula for the weighted summation operation of the standard AIS data according to the attention weights is as follows:

[0060]

[0061] Where softmax is the activation function, W is the attention weight, the dimension is (num_features, 1), num_features is the dimension of the standard AIS data, feature_t is the standard AIS data at time step t, N is the total number of time steps, and softmax(W*feature_t) represents the attention score.

[0062] Understandably, this invention uses attention weights to perform a weighted summation operation on the processed historical AIS data to obtain an attention-adjusted historical information representation, which better reflects important time windows and feature terms, thus helping to improve prediction performance. Combining the attention-adjusted historical information representation with the traditional CNN-LSTM model forms an end-to-end prediction system. The model will learn the key information of the processed historical AIS data more accurately, thereby improving the accuracy of flight trajectory prediction.

[0063] Based on the above embodiments, as an optional embodiment, the input data of the multi-scale CNN module consists of a three-dimensional tensor composed of data height, data width, and data channels, which can be represented as (height, width, channels), where height and width are the height and width of the input data, respectively, and channels are the number of channels in the input data. The multi-scale CNN module includes multiple convolutional layers, activation functions, pooling layers, and fully connected layers. Higher-level features are extracted by stacking and repeating multiple convolutional layers. The convolutional layers are used for convolution operations based on cross-correlation; the activation functions are used to perform nonlinear transformations on the feature maps output by the convolutional layers; the pooling layers are used to downsample the feature maps output by the convolutional layers based on max pooling. The output of the convolutional layers is passed through one or more fully connected layers, mapping the image features to the input of the classifier. Then, the output of the fully connected layers is flattened into a one-dimensional vector, which serves as the input to the LSTM module.

[0064] Optionally, the formula for implementing the convolution operation using cross-correlation in this invention is shown below:

[0065] output[row,col,filter]=Σ i Σ j Σ k input[row×stride+i,col×stride+j,k]×filter[i,j,k,filter];

[0066] Where output[row, col, filter] is the output value of the convolution operation, input[row×stride+i, col×stride+j, k] is the value of the input data, filter[i, j, k, filter] is the value of the convolution kernel, and stride represents the stride.

[0067] Optionally, this invention uses ReLU (Rectified Linear Unit) as the activation function, and the formula for the nonlinear transformation is as follows:

[0068] ReLU(x) = max(0, x)

[0069] Here, ReLU(x) is the activation function.

[0070] Optionally, the pooling layer's operation function is MaxPool(x) = max(x). i，j ).

[0071] Optionally, a multi-scale CNN module can obtain features at different scales by using convolutional kernels or pooling operations of varying sizes. Features extracted from different time scales can be fused using methods such as weighted summation, concatenation, and attention mechanisms to combine multi-scale features into a more comprehensive and richer feature representation. Inputting the fused multi-scale feature representation into an LSTM model for temporal modeling helps the model better learn and predict trends and patterns in flight trajectories. End-to-end training and validation optimize the parameters of the multi-scale feature fusion module and the overall model to achieve more accurate flight trajectory prediction performance.

[0072] Understandably, multi-scale CNN modules can help models better understand information at different time scales, thereby more comprehensively capturing and utilizing the complex features of fishing vessel navigation trajectory data. In fishing vessel navigation trajectory prediction tasks, features at different time scales often contain different information and patterns. By fusing these multi-scale features, the model can become more flexible and generalizable.

[0073] Based on the above embodiments, as an optional embodiment, the input data format of the LSTM module is (batch_size, timesteps, features), where batch_size represents the number of samples in each training batch, timesteps represents the number of time steps in the sequence, and features represents the number of features at each time step. The LSTM module includes three gates (input gate, forget gate, and output gate) and a memory unit.

[0074] The formula for calculating the input gate is: i t =σ(W xi ·x t +W hi ·h t-1 +b i ), where x t W is the input of the current time step in the input sequence. xi and W hi It is the weight matrix of the input gate, b i It is the bias of the input gate.

[0075] The forgetting gate filters out previous memories; the calculation formula is: f t =σ(Wxf ·x t +W hf ·h t-1 +b f ), where W xf and W hf This is the weight matrix of the forget gate, b f It is the bias of the forget gate.

[0076] The formula for updating memory units is: C t =f t ·C t-1 +i t ·g t , where C t-1 It is the memory unit value of the previous time step, g t It is a candidate memory value for the current time step.

[0077] Output gates are used to control the output of memory cells. The output gate formula is: o t =σ(W xo ·x t +W ho ·h t-1 +b o ), where W xo and W ho It is the weight matrix of the output gate, b o It is the bias of the output gate.

[0078] The output gates and memory cells are used to calculate the hidden state (output) of the LSTM. The formula for calculating the hidden state is: h t =o t ·tanh(C t ).

[0079] During the forward propagation process, the input at each time step is processed by the LSTM module, which then outputs the hidden state h. t These hidden states can be further used in other parts of the improved CNN-LSTM network model.

[0080] For each input time t, the LSTM module updates the hidden state h using the above formula. t and cell state C t The hidden state sequence (h1, h2, ..., h) T ) as output.

[0081] It is understood that this invention utilizes the LSTM module to process time series data, capture time features such as the movement trends and historical behavior patterns of fishing vessels, process and predict important events with relatively long intervals and delays in the time series, and fully consider the temporal relationships in the trajectory data to provide more accurate and reliable results for fishing vessel trajectory prediction.

[0082] Based on the above embodiments, as an optional embodiment, after obtaining the predicted trajectory of the fishing vessel to be predicted, steps S300 and S400 are further included.

[0083] Step S300: Obtain the operating environment data and current operating target of the fishing vessel to be predicted;

[0084] Step S400: Based on the operational environment data, the current operational target, and the predicted trajectory, perform trajectory anomaly detection on the fishing vessel to be predicted for operation. Specifically, a mapping relationship between all operational targets, operational trajectories, and operational environments can be pre-constructed. Based on the current operational target of the fishing vessel to be predicted and the mapping relationship, the target operational trajectory and target operational environment corresponding to the current operational target can be determined. By comparing the predicted trajectory and the target operational trajectory, and then comparing the operational environment data and the target operational environment, the trajectory of the fishing vessel to be predicted for operation can be judged based on the comparison results.

[0085] It is understood that the present invention can monitor fishing vessels by detecting trajectory anomalies in the fishing vessels to be predicted.

[0086] It should be noted that the execution subject of the fishing vessel navigation trajectory prediction method provided by the present invention can be a server, computer equipment, such as mobile phone, tablet computer, laptop computer, handheld computer, vehicle electronic equipment, wearable device, ultra-mobile personal computer (UMPC), netbook or personal digital assistant (PDA), etc.

[0087] The following describes the fishing vessel trajectory prediction device provided by the present invention. The fishing vessel trajectory prediction device described below and the fishing vessel trajectory prediction method described above can be referred to in correspondence.

[0088] Figure 2 This is a schematic diagram of the structure of the fishing vessel navigation trajectory prediction device provided by the present invention, as shown below. Figure 2 As shown, the present invention also provides a navigation trajectory prediction device for fishing vessels, including a training module 210, an acquisition module 220 and a prediction module 230.

[0089] Training module 210 is used to acquire historical AIS data of fishing vessels, obtain standard AIS data based on the historical AIS data, improve the CNN-LSTM model based on attention mechanism and multi-scale mechanism to obtain improved CNN-LSTM model, and train the improved CNN-LSTM model based on the standard AIS data to obtain the trajectory prediction model.

[0090] The acquisition module 220 is used to acquire AIS data of the fishing vessel to be predicted.

[0091] The prediction module 230 is used to input the AIS data into a pre-trained trajectory prediction model to obtain the predicted trajectory of the fishing vessel to be predicted.

[0092] As an example, the CNN-LSTM model includes a CNN module and an LSTM module connected to the output of the CNN module. The training module 210 is further configured to: set an attention module at the input of the CNN module, the attention module being used to determine the attention weights of the standard AIS data or the AIS data at each time step based on a content-based attention mechanism; perform a weighted summation operation on the standard AIS data according to the attention weights to obtain attention-adjusted standard AIS data; adjust the convolution kernel parameters or pooling layer parameters of the CNN module to obtain a multi-scale CNN module, the multi-scale CNN module being used to extract the spatial features of the attention-adjusted standard AIS data at different time scales and perform feature fusion to obtain multi-scale fused features; and obtain the improved CNN-LSTM model based on the attention module, the multi-scale CNN module, and the LSTM module, the LSTM module being used to extract the temporal features of the multi-scale fused features.

[0093] As an example, the calculation formula for the weighted summation operation of the standard AIS data based on the attention weights is as follows:

[0094]

[0095] Where softmax is the activation function, W is the attention weight, feature_t is the standard AIS data at time step t, and N is the total number of time steps.

[0096] As an example, the input data of the multi-scale CNN module consists of a three-dimensional tensor composed of data height, data width, and data channels. The multi-scale CNN module includes multiple convolutional layers, activation functions, pooling layers, and fully connected layers. The convolutional layers are used to perform convolution operations based on cross-correlation. The activation functions are used to perform nonlinear transformations on the feature maps output by the convolutional layers. The pooling layers are used to perform downsampling operations on the feature maps output by the convolutional layers based on max pooling.

[0097] As an example, the acquisition module 220 is further configured to: perform data cleaning on the historical AIS data; format and convert the cleaned historical AIS data according to a preset data format, wherein the preset data format is a unique ship identifier, direction, longitude, latitude, heading, and upload time; and normalize the formatted and converted historical AIS data to obtain the standard AIS data.

[0098] The formula for normalization is as follows:

[0099]

[0100] Where x represents historical AIS data. σ is the mean of the historical AIS data, and σ is the standard deviation of the historical AIS data.

[0101] As an example, the prediction module 230 is further configured to: acquire the operating environment data and current operating target of the fishing vessel to be predicted; and perform trajectory anomaly detection on the fishing vessel to be predicted based on the operating environment data, the current operating target, and the predicted trajectory.

[0102] It should be noted that the fishing vessel navigation trajectory prediction device provided by the present invention can execute the fishing vessel navigation trajectory prediction method described in any of the above embodiments during specific operation, and has the technical effects corresponding to the method. This embodiment will not elaborate on this.

[0103] Figure 3 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 3As shown, the electronic device may include: a processor 310, a communication interface 320, a memory 330, and a communication bus 340, wherein the processor 310, the communication interface 320, and the memory 330 communicate with each other through the communication bus 340. The processor 310 can call logical instructions in the memory 330 to execute a method for predicting the navigation trajectory of a fishing vessel. This method includes: acquiring AIS data of the fishing vessel to be predicted; inputting the AIS data into a pre-trained trajectory prediction model to obtain the predicted trajectory of the fishing vessel; wherein the training steps of the trajectory prediction model include: acquiring historical AIS data of the fishing vessel, obtaining standard AIS data based on the historical AIS data; improving the CNN-LSTM model based on attention and multi-scale mechanisms to obtain an improved CNN-LSTM model; and training the improved CNN-LSTM model based on the standard AIS data to obtain the trajectory prediction model.

[0104] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0105] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the fishing vessel trajectory prediction method provided by the above methods. The method includes: acquiring AIS data of the fishing vessel to be predicted; inputting the AIS data into a pre-trained trajectory prediction model to obtain the predicted trajectory of the fishing vessel; wherein the training step of the trajectory prediction model includes: acquiring historical AIS data of the fishing vessel, obtaining standard AIS data based on the historical AIS data; improving the CNN-LSTM model based on attention mechanism and multi-scale mechanism to obtain an improved CNN-LSTM model; and training the improved CNN-LSTM model based on the standard AIS data to obtain the trajectory prediction model.

[0106] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the method for predicting the navigation trajectory of a fishing vessel provided by the above methods. The method includes: acquiring AIS data of the fishing vessel to be predicted; inputting the AIS data into a pre-trained trajectory prediction model to obtain the predicted trajectory of the fishing vessel; wherein the training step of the trajectory prediction model includes: acquiring historical AIS data of the fishing vessel, obtaining standard AIS data based on the historical AIS data; improving the CNN-LSTM model based on attention and multi-scale mechanisms to obtain an improved CNN-LSTM model; and training the improved CNN-LSTM model based on the standard AIS data to obtain the trajectory prediction model.

[0107] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0108] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0109] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting the navigation trajectory of a fishing vessel, characterized in that, include: Obtain AIS data of the fishing vessels to be predicted for operation; Input the AIS data into the pre-trained trajectory prediction model to obtain the predicted trajectory of the fishing vessel to be predicted. The training steps of the trajectory prediction model include: Acquire historical AIS data of the fishing vessel in operation, and obtain standard AIS data based on the historical AIS data; An improved CNN-LSTM model is obtained by improving the attention mechanism and the multi-scale mechanism. The improved CNN-LSTM model is trained based on the standard AIS data to obtain the trajectory prediction model; The CNN-LSTM model includes a CNN module and an LSTM module connected to the output of the CNN module. The improved CNN-LSTM model, based on attention and multi-scale mechanisms, includes: An attention module is set at the input of the CNN module. The attention module is used to determine the attention weight of the standard AIS data at each time step using a content-based attention mechanism. The standard AIS data is then weighted and summed according to the attention weight to obtain the attention-adjusted standard AIS data. The content-based attention mechanism is a soft attention mechanism, which adjusts the weight of the time step according to the fishing vessel's operation target or operation mode. Adjusting the convolution kernel parameters or pooling layer parameters of the CNN module yields a multi-scale CNN module. The multi-scale CNN module is used to extract spatial features of the attention-adjusted standard AIS data at different time scales and perform feature fusion to obtain multi-scale fused features. Based on the attention module, the multi-scale CNN module, and the LSTM module, the improved CNN-LSTM model is obtained, wherein the LSTM module is used to extract the temporal features of the multi-scale fusion features; The calculation formula for the weighted summation operation of the standard AIS data based on the attention weight is as follows: ； in, For activation function, For attention weights, For the first t Standard AIS data at each time step, N This represents the total number of time steps. The input data of the multi-scale CNN module consists of a three-dimensional tensor composed of data height, data width, and data channels. The multi-scale CNN module includes multiple convolutional layers, activation functions, pooling layers, and fully connected layers. The convolutional layers are used to perform convolution operations based on cross-correlation; the activation functions are used to perform nonlinear transformations on the feature maps output by the convolutional layers; and the pooling layers are used to perform downsampling operations on the feature maps output by the convolutional layers based on max pooling. After obtaining the predicted trajectory of the fishing vessel to be predicted, the process further includes: Obtain the operational environment data and current operational target of the fishing vessel to be predicted; Based on the operational environment data, the current operational target, and the predicted trajectory, trajectory anomaly detection is performed on the fishing vessel to be predicted. A mapping relationship between all operational targets, operational trajectories, and operational environments is pre-constructed. Based on the current operational target and mapping relationship of the fishing vessel to be predicted, the target operational trajectory and target operational environment corresponding to the current operational target are determined. By comparing the predicted trajectory and the target operational trajectory, and then comparing the operational environment data and the target operational environment, the trajectory of the fishing vessel to be predicted is judged to be normal based on the comparison results.

2. The method for predicting the navigation trajectory of fishing vessels according to claim 1, characterized in that, The process of obtaining standard AIS data based on the historical AIS data includes: Perform data cleaning on the historical AIS data; The historical AIS data after cleaning is formatted and converted based on a preset data format, which includes the ship's unique identifier, voyage, longitude, latitude, heading, and upload time. The formatted and converted historical AIS data is normalized to obtain the standard AIS data; The formula for normalization is as follows: ； in, It is historical AIS data. It is the average of historical AIS data. It is the standard deviation of historical AIS data.

3. A device for predicting the navigation trajectory of a fishing vessel, characterized in that, The method for predicting the navigation trajectory of a fishing vessel according to any one of claims 1-2 includes: The training module is used to acquire historical AIS data of fishing vessels and obtain standard AIS data based on the historical AIS data; improve the CNN-LSTM model based on the attention mechanism and multi-scale mechanism to obtain an improved CNN-LSTM model; train the improved CNN-LSTM model based on the standard AIS data to obtain the trajectory prediction model. The acquisition module is used to acquire AIS data of the fishing vessels to be predicted. The prediction module is used to input the AIS data into a pre-trained trajectory prediction model to obtain the predicted trajectory of the fishing vessel to be predicted.

4. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the fishing vessel navigation trajectory prediction method as described in any one of claims 1 to 2.

5. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method for predicting the navigation trajectory of a fishing vessel as described in any one of claims 1 to 2.

6. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the method for predicting the navigation trajectory of a fishing vessel as described in any one of claims 1 to 2.

Images