Method and system for predicting drifting trajectory of offshore personnel falling into water

Abstract

The invention relates to the technical field of drifting trajectory prediction of marine targets, in particular to a drifting trajectory prediction method and a drifting trajectory prediction system for marine personnel falling into water. Training set data and test set data are read from drifting data, and time serialization processing operation is performed on the training set data and the test set data; the method comprises the steps that training set time sequence data and test set time sequence data are obtained, the training set time sequence data and the test set time sequence data are substituted into an LSTM model to obtain a training model, the test set data are tested through the training model to obtain a trajectory prediction model, and the LSTM model can capture a long-term dependency relationship in the time sequence data. The method is suitable for various sequence prediction tasks, so that the real drifting situation of people falling into water under different sea conditions can be accurately simulated.

Description

Technical Field

[0001] The present invention relates to the technical field of drift trajectory prediction of marine targets, and in particular to a method and system for predicting the drift trajectory of a person falling into water at sea. Background Art

[0002] With the expansion of the scale of marine development and the frequent maritime activities, maritime accidents have attracted more attention. Maritime accidents will bring huge disasters to coastal countries in the world in terms of politics, economy and military. Not only will they cause casualties and property losses, but they will also have adverse effects on social development. Therefore, maritime search and rescue work has received more and more attention from coastal countries. For the rapidly developing maritime transportation and fishing industries, maritime search and rescue operations can provide irreplaceable protection for the safety of personnel and property.

[0003] At present, my country's maritime search and rescue command and coordination work mainly relies on the experience and judgment of the search and rescue commanders. It is not possible to quickly and efficiently predict the drift trajectory of the accident victims based on the hydrological and meteorological conditions at the accident site at sea. This has greatly affected the speed and accuracy of the command and coordination work, and often caused the search and rescue operations to miss opportunities. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a method and system for predicting the drift trajectory of a person falling into water at sea, which can accurately simulate the actual situation of drifting after a person falls into water under different sea conditions.

[0005] In order to solve the above technical problems, the first technical solution adopted by the present invention is:

[0006] A method for predicting the drift trajectory of a person falling into water at sea comprises the following steps:

[0007] S1. Obtain drift data;

[0008] S2, read the training set data and the test set data from the drift data;

[0009] S3, performing a time series processing operation on the training set data and the test set data to obtain training set time series data and test set time series data;

[0010] S4. Build LSTM model;

[0011] S5, substituting the training set time series data and the test set time series data into the LSTM model to obtain a training model;

[0012] S6. Test the test set data through the training model to obtain the trajectory prediction model.

[0013] The second technical solution adopted by the present invention is:

[0014] A system for predicting drifting trajectories of people falling into water at sea includes one or more processors and a memory, wherein the memory stores a program, and when the program is executed by the processor, the following steps are implemented:

[0015] S1. Obtain drift data;

[0016] S2, read the training set data and the test set data from the drift data;

[0017] S3, performing a time series processing operation on the training set data and the test set data to obtain training set time series data and test set time series data;

[0018] S4. Build LSTM model;

[0019] S5, substituting the training set time series data and the test set time series data into the LSTM model to obtain a training model;

[0020] S6. Test the test set data through the training model to obtain the trajectory prediction model.

[0021] The beneficial effects of the present invention are:

[0022] This scheme reads the training set data and the test set data from the drifting data, performs time serialization processing on the training set data and the test set data to obtain the training set time series data and the test set time series data, substitutes the training set time series data and the test set time series data into the LSTM model to obtain the training model, and tests the test set data through the training model to obtain the trajectory prediction model. The LSTM model can capture the long-term dependencies in the time series data and is suitable for various sequence prediction tasks, so as to accurately simulate the actual situation of drifting after people fall into the water under different sea conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 A flowchart of the steps of the method for predicting the drift trajectory of a person falling into water at sea according to the present invention;

[0024] Figure 2 A network structure diagram of an LSTM model of a method for predicting drift trajectories of people falling into water at sea according to the present invention;

[0025] Figure 3 It is a structural block diagram of the drift trajectory prediction system for people falling into water at sea of ​​the present invention;

[0026] Description of labels:

[0027] 1. Processor; 2. Memory. DETAILED DESCRIPTION

[0028] In order to explain the technical content, achieved objectives and effects of the present invention in detail, the following is an explanation in combination with the implementation modes and the accompanying drawings.

[0029] Please refer to Figure 1 , the first technical solution adopted by the present invention is:

[0030] A method for predicting the drift trajectory of a person falling into water at sea comprises the following steps:

[0031] S1. Obtain drift data;

[0032] S2, read the training set data and the test set data from the drift data;

[0033] S3, performing a time series processing operation on the training set data and the test set data to obtain training set time series data and test set time series data;

[0034] S4. Build LSTM model;

[0035] S5, substituting the training set time series data and the test set time series data into the LSTM model to obtain a training model;

[0036] S6. Test the test set data through the training model to obtain the trajectory prediction model.

[0037] From the above description, it can be seen that the beneficial effects of the present invention are:

[0038] This scheme reads the training set data and the test set data from the drifting data, performs time serialization processing on the training set data and the test set data to obtain the training set time series data and the test set time series data, substitutes the training set time series data and the test set time series data into the LSTM model to obtain the training model, and tests the test set data through the training model to obtain the trajectory prediction model. The LSTM model can capture the long-term dependencies in the time series data and is suitable for various sequence prediction tasks, so as to accurately simulate the actual situation of drifting after people fall into the water under different sea conditions.

[0039] Furthermore, after step S6, the following steps are also included:

[0040] S7, obtaining predicted trajectory data according to the trajectory prediction model;

[0041] S8. Substituting the predicted trajectory data into a Kalman filter to obtain an optimized trajectory prediction model.

[0042] From the above description, it can be seen that the Kalman filter can further optimize the trajectory prediction model and provide a more accurate predicted trajectory.

[0043] Furthermore, the following steps are included between step S2 and step S3:

[0044] A normalization operation is performed on the training set data and the test set data.

[0045] From the above description, we can see that by normalizing the training set data and the test set data, the model can better learn the relationship between features.

[0046] Furthermore, step S1 specifically includes the following steps:

[0047] Throw the float at the designated location, use the positioning module on the float to perform real-time positioning, and record the drift data of the float.

[0048] Please refer to Figure 3 , the second technical solution adopted by the present invention is:

[0049] A system for predicting drifting trajectories of people falling into water at sea includes one or more processors and a memory, wherein the memory stores a program, and when the program is executed by the processor, the following steps are implemented:

[0050] S1. Obtain drift data;

[0051] S2, read the training set data and the test set data from the drift data;

[0052] S3, performing a time series processing operation on the training set data and the test set data to obtain training set time series data and test set time series data;

[0053] S4. Build LSTM model;

[0054] S5, substituting the training set time series data and the test set time series data into the LSTM model to obtain a training model;

[0055] S6. Test the test set data through the training model to obtain the trajectory prediction model.

[0056] Furthermore, when the program is executed by the processor, the following steps are implemented:

[0057] S7, obtaining predicted trajectory data according to the trajectory prediction model;

[0058] S8. Substituting the predicted trajectory data into a Kalman filter to obtain an optimized trajectory prediction model.

[0059] Furthermore, when the program is executed by the processor, the following steps are implemented:

[0060] A normalization operation is performed on the training set data and the test set data.

[0061] Furthermore, when the program is executed by the processor, the following steps are implemented:

[0062] Throw the float at the designated location, use the positioning module on the float to perform real-time positioning, and record the drift data of the float.

[0063] Please refer to Figure 1 and Figure 2 As shown, the first embodiment of the present invention is:

[0064] Please refer to Figure 1 , a method for predicting the drift trajectory of a person falling into water at sea, comprising the following steps:

[0065] S1, obtaining drift data; Step S1 specifically includes the following steps:

[0066] Throw the float at the designated location, use the positioning module on the float to locate it in real time, and record the drift data of the float;

[0067] S2. Read the training set data and the test set data from the drift data; use the pd.read_excel function to read the training set data and the test set data; use the float data set numbered 17 as the training set data and the float data set numbered 20 as the test set data;

[0068] S3, the training set data and the test set data are subjected to time series processing to obtain the training set time series data and the test set time series data; the time step is set to 3 through the variable time_steps (referring to the time step) in the code, and the model will use the data of the past three time points to predict the value of the next time point; two for loop functions are used to perform time series processing on the training set data and the test set data respectively; in each loop, continuous time_steps data points are extracted through slicing operations, and they are stored as a sequence in the list X_train_sequence or X_test_sequence (referring to the training set); at the same time, the first data point after the sequence is stored as the target value in y_train_next_step or y_test_next_step (referring to the test set); the lists X_train_sequence and X_test_sequence are converted into NumPy arrays for model training and testing;

[0069] The following steps are also included between step S2 and step S3:

[0070] The training set data and the test set data are normalized; first, the MinMaxScaler class function is imported, a normalization object is created through the MinMaxScaler function, and the training set feature data X_train is fitted and transformed using the scaler.fit_transform(X_train) function; this step calculates the minimum and maximum values ​​of the training set data, and transforms the data to a range of 0 to 1 based on these values; the test set feature data X_test is transformed to the same scale as the training set using the scaler.transform(X_test) function; the transformed data is stored in new variables X_train_scaled and X_test_scaled (referring to the training set scale) for subsequent use in model training and testing;

[0071] S4. Build LSTM model;

[0072] The LSTM (Long Short-Term Memory) model is a special type of recurrent neural network that is suitable for processing time series data. It can learn long-term dependencies in time series data. Compared with traditional RNN (Recurrent Neural Network), LSTM solves the problems of gradient vanishing and gradient exploding by introducing a gating mechanism, enabling it to process longer sequence data.

[0073] The LSTM network consists of multiple LSTM units, each of which contains three gates: Forget Gate, Input Gate, and Output Gate. Figure 2 shown.

[0074] The forget gate determines what information in the cell state should be forgotten or discarded. It contains a sigmoid activation function that outputs a value between 0 and 1 based on the input data and the hidden state of the previous time step.

[0075] The input gate controls how much the new input information affects the current cell state; it includes a sigmoid activation function that generates a value between 0 and 1 indicating which values ​​in each cell state should be updated.

[0076] The output gate controls how the hidden state at the current moment and the cell state at the next moment affect the final output; it contains a sigmoid activation function to determine which parts of the output state will be activated.

[0077] These gate structures enable LSTM to selectively retain or forget information during training, thereby effectively solving the problems of gradient vanishing and gradient exploding. Specifically, the forget gate is responsible for deciding which information to discard from the cell state, the input gate is responsible for deciding which new information to add to the cell state, and the output gate is responsible for deciding which information to output from the cell state. Through this mechanism, LSTM is able to capture long-term dependencies in time series data and is suitable for various sequence prediction tasks, such as language models, time series prediction, and bioinformatics.

[0078] Use the fit function to train the model. This function accepts the time-series training set data X_train_sequence and target data y_train_next_step, as well as the validation data X_test_sequence and y_test_next_step. During the training process, the LSTM model will learn how to predict the value of the next time point based on the input time series data. In the fit function, the LSTM model will be trained for multiple epochs (iterations). In each epoch, the LSTM model will traverse the entire training set data and update the model weights according to the loss function. This process is implemented through the backpropagation algorithm.

[0079] S5, substituting the training set time series data and the test set time series data into the LSTM model to obtain a training model;

[0080] S6. Test the test set data through the training model to obtain the trajectory prediction model.

[0081] After step S6, the following steps are also included:

[0082] S7, obtaining predicted trajectory data according to the trajectory prediction model;

[0083] S8. Substituting the predicted trajectory data into a Kalman filter to obtain an optimized trajectory prediction model.

[0084] First, the test data needs to go through the same preprocessing steps as the training set data, including feature selection, normalization, etc. The test data also needs to be converted to a time series format. This is done in the same way as the training set data, that is, slicing the data into a sequence of time_steps consecutive time points. Use the trained model to predict the time series test data. The trained model will output the predicted value for the next time point in each sequence. The performance of the model is evaluated by comparing the predicted values ​​of the model with the actual values ​​in the test data. The mean square error (MSE) and the coefficient of determination (R) can be used to 2 As an evaluation indicator; finally, the actual value and predicted value are visualized to intuitively show the prediction effect of the model.

[0085] The core of the Kalman filter lies in its recursive estimation process, which includes a prediction phase and an update phase; through the two steps of prediction and update, prior knowledge and observation data are effectively integrated to estimate the system state.

[0086] In the prediction step (time update), the Kalman filter predicts the state at the current time based on the state estimate and control input at the previous time and estimates the covariance of the prediction error.

[0087] Predicted state equation:

[0088] Prediction covariance equation:

[0089] In the above formula, is the state estimate at time k; A is the state transfer matrix, which describes the transition of the system state; B is the control input matrix, which describes the influence of the control input on the state; u k is the control input vector; Q is the process noise covariance matrix, which describes the unpredictable part of the system model;

[0090] In the measurement update step, the Kalman filter combines the observed data and the prediction error to adjust the predicted state through the Kalman gain to obtain a more accurate estimate.

[0091] Kalman gain equation:

[0092] The updated prediction state equation is:

[0093] Updated forecast covariance equation:

[0094] In the above formula, the state vector x k =Ax k-1 +Bu k +w k , where u k is the control input vector, w k is the process noise, usually assumed to be Gaussian white noise with zero mean and covariance Q.

[0095] Measurement Model z k =Hx k +v k , where H is the measurement matrix, v k is the measurement noise, which is usually assumed to be Gaussian white noise with zero mean and covariance R.

[0096] The Kalman filter weighs past state estimates and current measurements to minimize the mean square error of the estimation error to obtain the optimal estimate of the system state; it achieves continuous tracking and adjustment of the system state by recursively performing predictions and measurement updates.

[0097] The predicted trajectory data generated by the trajectory prediction model is used as the initial prediction of the Kalman filter, and the measurement value (i.e. the trajectory predicted by the trajectory prediction model) and the noise covariance (measurement noise covariance R and process noise covariance Q) are used as input to optimize the trajectory estimation through the prediction step and the update step. Initial state estimate Use the position of the last time step predicted by the trajectory prediction model as the initial position estimate, and use the identity matrix multiplied by 0.4 as the initial covariance P0.

[0098] Calculate the mean square error MSE and determination coefficient R between the Kalman filter output and the trajectory prediction model direct output 2 , Please refer to Table 1 for specific experimental data.

[0099] Mean square error (MSE): It is used to measure the average value of the squared difference between the predicted value and the actual value. The smaller the value, the more accurate the model prediction.

[0100]

[0101] Where n is the number of samples, y i is the ith actual value, is the i-th predicted value.

[0102] Coefficient of determination R 2 : A statistic that measures the model's fit to the data, between 0 and 1. The closer it is to 1, the better the model fit.

[0103]

[0104] Where n is the number of samples, y i is the ith actual value, is the ith predicted value, is the average of the actual values.

[0105] The specific simulation results are compared with the experimental results as follows:

[0106]

[0107] Table 1

[0108] As can be seen from Table 1, the trajectory prediction after Kalman filter post-processing has a lower mean square error MSE and a higher determination coefficient R 2, which shows that the Kalman filter can effectively reduce the prediction error and improve the accuracy of trajectory prediction.

[0109] In order to further improve the predictive ability of the model, the following improvements can also be considered:

[0110] Adjust the hyperparameters of the LSTM model, such as increasing the number of LSTM units and adjusting the learning rate, to explore a more suitable model structure; the application of the Kalman filter improves the prediction performance of the LSTM model, but its performance is also limited by the selection of filter parameters and the accuracy of the system model; future work will focus on further optimizing the Kalman filter parameters and exploring the combination of other filtering techniques with the LSTM model to achieve more accurate prediction of the drift trajectory of people falling into the water at sea; use more features or introduce additional environmental variables (such as marine environmental parameters) to enrich the input information of the model; perform more detailed analysis and preprocessing of the data set to improve the model's understanding and capture of the buoy's drift characteristics.

[0111] These improvements can help optimize the performance of the model, enabling it to more accurately predict the drift trajectory of falling water under different sea conditions, and improve the reliability and effectiveness of practical applications.

[0112] Please refer to Figure 3 , Embodiment 2 of the present invention is:

[0113] A system for predicting drifting trajectories of people falling into water at sea includes one or more processors 1 and a memory 2, wherein the memory 2 stores a program, and when the program is executed by the processor 1, the following steps are implemented:

[0114] S1. Obtain drift data;

[0115] S2, read the training set data and the test set data from the drift data;

[0116] S3, performing a time series processing operation on the training set data and the test set data to obtain training set time series data and test set time series data;

[0117] S4. Build LSTM model;

[0118] S5, substituting the training set time series data and the test set time series data into the LSTM model to obtain a training model;

[0119] S6. Test the test set data through the training model to obtain the trajectory prediction model.

[0120] When the program is executed by processor 1, the following steps are implemented:

[0121] S7, obtaining predicted trajectory data according to the trajectory prediction model;

[0122] S8. Substituting the predicted trajectory data into a Kalman filter to obtain an optimized trajectory prediction model.

[0123] When the program is executed by processor 1, the following steps are implemented:

[0124] A normalization operation is performed on the training set data and the test set data.

[0125] When the program is executed by processor 1, the following steps are implemented:

[0126] Throw the float at the designated location, use the positioning module on the float to perform real-time positioning, and record the drift data of the float.

[0127] In summary, the present invention provides a method and system for predicting the drift trajectory of a person falling into the water at sea. The method and system read the training set data and the test set data from the drift data, perform time series processing operations on the training set data and the test set data, obtain the training set time series data and the test set time series data, substitute the training set time series data and the test set time series data into the LSTM model to obtain the training model, test the test set data through the training model, and obtain the trajectory prediction model. The LSTM model can capture the long-term dependencies in the time series data and is suitable for various sequence prediction tasks, so as to accurately simulate the actual situation of drifting after a person falls into the water under different sea conditions.

[0128] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent transformations made using the contents of the present invention's specification and drawings, or directly or indirectly applied in related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for predicting the drift trajectory of a person falling into water at sea, characterized in that: The following steps are involved: S1. Obtain drift data; S2, read the training set data and the test set data from the drift data; S3, performing a time series processing operation on the training set data and the test set data to obtain training set time series data and test set time series data; S4. Build LSTM model; S5, substituting the training set time series data and the test set time series data into the LSTM model to obtain a training model; S6. Test the test set data through the training model to obtain the trajectory prediction model.

2. The method for predicting the drift trajectory of a person falling into water at sea according to claim 1, characterized in that: After step S6, the following steps are also included: S7, obtaining predicted trajectory data according to the trajectory prediction model; S8. Substituting the predicted trajectory data into a Kalman filter to obtain an optimized trajectory prediction model.

3. The method for predicting the drift trajectory of a person falling into water at sea according to claim 1, characterized in that: The following steps are also included between step S2 and step S3: A normalization operation is performed on the training set data and the test set data.

4. The method for predicting the drift trajectory of a person falling into water at sea according to claim 1, characterized in that: Step S1 specifically includes the following steps: Throw the float at the designated location, use the positioning module on the float to perform real-time positioning, and record the drift data of the float.

5. A system for predicting the drift trajectory of a person falling into water at sea, characterized in that: The system comprises one or more processors and a memory, wherein the memory stores a program, and when the program is executed by the processor, the following steps are implemented: S1. Obtain drift data; S2, read the training set data and the test set data from the drift data; S3, performing a time series processing operation on the training set data and the test set data to obtain training set time series data and test set time series data; S4. Build LSTM model; S5, substituting the training set time series data and the test set time series data into the LSTM model to obtain a training model; S6. Test the test set data through the training model to obtain the trajectory prediction model.

6. The system for predicting drifting trajectory of a person falling into water at sea according to claim 5, characterized in that: When the program is executed by the processor, the following steps are implemented: S7, obtaining predicted trajectory data according to the trajectory prediction model; S8. Substituting the predicted trajectory data into a Kalman filter to obtain an optimized trajectory prediction model.

7. The system for predicting drifting trajectory of a person falling into water at sea according to claim 5, characterized in that: When the program is executed by the processor, the following steps are implemented: A normalization operation is performed on the training set data and the test set data.

8. The system for predicting drifting trajectory of a person falling into water at sea according to claim 5, characterized in that: When the program is executed by the processor, the following steps are implemented: Throw the float at the designated location, use the positioning module on the float to perform real-time positioning, and record the drift data of the float.

Images