Track fusion method and system based on Kalman filtering and convex combination theory
By combining Kalman filtering and convex combination theory in track fusion, the problems of high computational complexity, strong model dependence and low robustness in the prior art are solved, and more efficient and accurate track fusion is achieved.
Patent Information
- Application Number
- CN202411751635.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-02
- Publication Date
- 2025-05-06
AI Technical Summary
Existing track fusion algorithms such as particle filters and JPDA algorithms have problems such as high computational complexity, high dependence on model accuracy, low robustness, and experience in weight selection, which affects the accuracy and efficiency of track tracking.
The track fusion method based on Kalman filtering and convex combination theory is adopted to process the track data through the Kalman filtering algorithm, and the convex combination algorithm is used to fusion of multi-sensor data, simplifying the correlation problem and improving the accuracy of track correlation.
It improves the accuracy and efficiency of track fusion, reduces the computational complexity, enhances the robustness of model changes and abnormal data, and improves the operability of the algorithm through clear weight determination and data normalization steps.
Smart Images
Figure CN119939494A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of track fusion, and in particular relates to a track fusion method and system based on Kalman filtering and convex combination theory. Background Art
[0002] The track fusion algorithm based on particle filter simulates a group of particles representing the possible states of the target, updates the weights of these particles according to the sensor observation data, and realizes accurate estimation of the target track. It can effectively handle nonlinear and non-Gaussian systems, adapt to multi-sensor data fusion, and handle model uncertainty. The track fusion algorithm based on joint probabilistic data association (JPDA) calculates the joint probability between multiple sensor observations and multiple target trajectories to achieve accurate fusion of tracks. It can effectively handle data association problems in multi-target environments and improve the accuracy and robustness of track tracking.
[0003] Kalman filtering is an algorithm that uses the linear system state equation and the system input and output observation data to optimally estimate the system state. Kalman filtering uses the state equation and observation equation, combined with the system's dynamic model and observation data, to optimally estimate the system state in a recursive manner.
[0004] However, particle filters require a large number of particles to maintain the accuracy of estimation, which may lead to increased computational complexity and memory requirements. At the same time, the particle degeneration problem may lead to a decrease in particle diversity, affecting the performance of the algorithm.
[0005] The computational complexity of the JPDA algorithm is high, especially in high target density environments, and the processing speed may be limited. At the same time, the dependence on prior probability and noise model may lead to performance degradation, especially in the case of high model uncertainty.
[0006] The Kalman filter is highly dependent on the accuracy of the system dynamic model and the observation noise model. If the model estimation is inaccurate, the performance of the filter may be affected. In addition, this method is sensitive to model changes and abnormal data, and its robustness is relatively low.
[0007] The convex combination fusion algorithm needs to perform weighted averaging of the outputs of different algorithms, which may require additional calculations and parameter adjustments. At the same time, the selection and adjustment of weights may depend on experience, which may affect the performance fluctuations of the algorithm.
[0008] In view of this, it is very meaningful to propose a track fusion method and system based on Kalman filtering and convex combination theory. Summary of the invention
[0009] In order to solve the problems existing in the above-mentioned technologies, the present invention provides a track fusion method and system based on Kalman filtering and convex combination theory to solve the above-mentioned technical defects.
[0010] In a first aspect, the present invention proposes a track fusion method based on Kalman filtering and convex combination theory, the method comprising the following steps:
[0011] Data collection: collect ship AIS data information collected by a base station in a certain place;
[0012] Data processing, data cleaning, data conversion and data aggregation processing of the collected ship AIS data information;
[0013] Track data generation: Assuming that the ship's trajectory is a uniform linear motion, the track data is generated based on the object motion expression, taking into account the loss caused by external influences to obtain the observation model;
[0014] Kalman filter processing, using the Kalman filter algorithm to process the generated track data, the Kalman filter algorithm consists of a time update equation and a measurement update equation, and processes the track information of each sensor in turn according to a predetermined order until the prediction step of the track information of all sensors is completed;
[0015] Convex combination fusion, input the predicted trajectory information of each sensor at the end of the Kalman filter iteration, and use the convex combination algorithm to perform track fusion.
[0016] Preferably, the track data generation specifically includes:
[0017] Assuming that the ship's trajectory is uniform linear motion, the object's motion expression is s = s0 + vt, where s represents the object's position at a certain moment, s0 is the object's initial position, v represents the object's speed, and t is time;
[0018] Expressed in matrix form: x = [0 t][x0 x] T , x is a vector representation related to the state of motion of the object, x0 is the initial position of the object; it is expressed as a state variable: x = [0 t][x k-1 x k-1 ] T =F k-1 x k-1 , x k-1 represents the state variable of the object at the k-1th moment, F k-1 is the state transfer matrix, which describes the transfer law of the object state from the k-1th moment to the next moment;
[0019] The loss caused by external influence is: x k =[0 t][x k-1x k-1 ] T +[0 t]w k =F k-1 x k-1 +G k-1 w k-1 , x k represents the state variable of the object at the Kth moment, w k Represents external influencing factors, G k-1 is a matrix related to external influences, which describes the external influence factors w k-1 How to act on the state variables of an object;
[0020] The observation model is obtained as: C(k) is the observation model, which describes how to obtain the observation value from the actual state of the object, taking into account the state of external influences.
[0021] Preferably, the Kalman filter algorithm consists of a time update equation and a measurement update equation, and the time update equation is:
[0022]
[0023] in, represents the predicted value of the system state at time k based on the observation information at time k-1 and before, F k-1 is the state transfer matrix, which describes the transition law of the system state from time k-1 to time k. is the estimated value of the system state at time k based on the observation information at time k-1, P k|k-1 G represents the covariance matrix of the system state prediction value at time k based on the observation information at time k-1 and before, k-1 The matrix related to system noise describes how system noise affects the transition of system state, Q k-1 is the covariance matrix of the system noise, which is used to describe the statistical characteristics of the system noise;
[0024] The measurement update equation is:
[0025]
[0026] Among them, K k is the Kalman gain matrix, which is used to weigh the predicted value and the observed value in the measurement update process. k is the observation matrix, which describes the linear relationship between the system state and the observation value, Z k is the observed value at time k, R k is the covariance matrix of the observation noise, which is used to describe the statistical characteristics of the observation noise. represents the updated estimate of the system state at time k based on the observation information at time k, P k|k Represents the covariance matrix of the updated estimate of the system state at time k based on the observation information at time k.
[0027] Further preferably, the Kalman filter processing uses a Kalman filter algorithm to process the generated track data, the Kalman filter algorithm consists of a time update equation and a measurement update equation, and processes the track information of each sensor in turn in a predetermined order until the prediction step of the track information of all sensors is completed, specifically including:
[0028] Input the pre-processed data, i.e. the coordinate information of the ship;
[0029] Define the initial values of the system's state variables such as position, velocity, and observation data, and set the initial state estimates and covariance matrix: In the formula represents the estimated value of the system state based on the initial information at the initial time (k = 0), E[x(0)] is the mathematical expectation of the system at the initial time x(0), p(0|0) represents the covariance matrix of the estimated value of the system state at the initial time (k = 0), and x(0) is the actual state of the system at the initial time;
[0030] Use the system's state transition equation to predict the state at the next moment and update the covariance matrix of the predicted state;
[0031] The Kalman gain is calculated using the covariance matrix of the predicted state and the covariance matrix of the observation noise;
[0032] Use the Kalman gain and the observation value to update the state estimate and the covariance matrix of the state estimate, use the updated state estimate and covariance matrix as the initial value of the next iteration, and repeat the prediction and update steps;
[0033] Determine whether the end condition is met, that is, whether the track information prediction of all sensors is completed.
[0034] Preferably, the algorithm formula of the convex combination fusion is:
[0035]
[0036] in, represents the estimated value of the system state after fusion, N is the number of sensors, P i is the state estimation covariance matrix corresponding to the i-th sensor, (P i ) -1 is the inverse matrix of the i-th sensor state estimation covariance matrix, is the estimated value of the system state by the i-th sensor; P -1 is the inverse matrix of the system state estimation covariance matrix after fusion, and P is the system state estimation covariance matrix after fusion.
[0037] Further preferably, the convex combination fusion inputs the predicted trajectory information of each sensor after the Kalman filter iteration ends, and uses the convex combination algorithm to perform track fusion, specifically including:
[0038] Input the predicted trajectory information of each sensor at the end of the Kalman filter iteration;
[0039] Initialize the sensor weights;
[0040] Based on the set weights, each predicted trajectory is weighted summed, and multiple inputs are fused into one output through linear combination;
[0041] Based on the evaluation results, feedback is provided to adjust the weights and parameters of the fusion strategy to optimize the fusion effect.
[0042] Preferably, the data processing includes data cleaning, data conversion and data aggregation processing on the collected ship AIS data information, specifically including:
[0043] Data cleaning: clean the collected ship AIS data, use preset rules and algorithms to accurately identify and remove invalid, missing or erroneous data, and delete incomplete and useless data;
[0044] Data conversion: converting the information contained in the cleaned AIS data into a data format that can be used for analysis, such as converting a timestamp into a specific date and time format;
[0045] Data aggregation: Aggregate the converted data and use a specific algorithm to aggregate the longitude and latitude position information of each ship into ship track information, and mark the ship track on the GIS chart.
[0046] In a second aspect, an embodiment of the present invention provides a track fusion system based on Kalman filtering and convex combination theory, including:
[0047] A data collection module is configured to collect ship AIS data information collected by a base station at a certain location;
[0048] A data processing module is configured to perform data cleaning, data conversion and data aggregation processing on the collected ship AIS data information;
[0049] The track data generation module is configured to assume that the ship's trajectory is a uniform linear motion, and to generate track data based on the object motion expression, wherein the loss caused by external influences is taken into account to obtain an observation model.
[0050] In a third aspect, an embodiment of the present invention provides an electronic device, comprising: one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any implementation manner in the first aspect.
[0051] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in any implementation manner in the first aspect.
[0052] Compared with the prior art, the beneficial results of the present invention are:
[0053] The sensor-to-system track fusion processing structure is adopted to send all data to the central system for unified fusion processing, simplifying the association problem and making it a bi-partite allocation problem. The common allocation algorithm can be used to avoid the inefficiency problem that may be caused by the discarding of past processing results in the sensor-to-sensor track fusion structure;
[0054] The use of Kalman filtering for track association has relatively loose requirements on the statistical properties of system disturbances and observation errors, and does not require the assumption that the signal and noise must be a stationary process. At the same time, it can process the observed signal containing noise to obtain an estimate of the real signal with the minimum error, thereby improving the accuracy of track association;
[0055] The convex combination fusion algorithm is used. Based on the convex combination theory, the results of multiple data sources are fused by weighted average, making the results more accurate. At the same time, the algorithm clarifies the method of determining the data source and its weight, as well as the specific steps of data normalization, convex combination calculation and anti-normalization, which has stronger operability and accuracy.
[0056] The technical solution of the present invention is technologically advanced in terms of processing structure, track association and track state estimation fusion, and can improve the effect and accuracy of track fusion. BRIEF DESCRIPTION OF THE DRAWINGS
[0057] The accompanying drawings are included to provide a further understanding of the embodiments and are incorporated into and constitute a part of this specification. The accompanying drawings illustrate the embodiments and are used together with the description to explain the principles of the present invention. It will be easy to recognize other embodiments and many expected advantages of the embodiments because they become better understood by reference to the following detailed description. The elements of the drawings are not necessarily to scale with each other. The same reference numerals refer to corresponding similar parts.
[0058] Figure 1A schematic flow chart of a track fusion method based on Kalman filtering and convex combination theory according to an embodiment of the present invention;
[0059] Figure 2 A schematic diagram of a track AIS data set collected according to an embodiment of the present invention;
[0060] Figure 3 This is a schematic diagram of a preprocessed track AIS data set according to an embodiment of the present invention;
[0061] Figure 4 A schematic diagram of sensor-to-system track fusion according to an embodiment of the present invention;
[0062] Figure 5 A schematic diagram of a Kalman filter mathematical model according to an embodiment of the present invention;
[0063] Figure 6 (a) and Figure 6 (b) are schematic diagrams of X-coordinate position tracking error and Y-coordinate position tracking error of an embodiment of the present invention;
[0064] Figure 7 is a schematic diagram of position error of an embodiment of the present invention;
[0065] Figure 8 A schematic diagram of aggregated ship trajectories according to an embodiment of the present invention;
[0066] Fig. 9 It is a schematic diagram of a track fusion system based on Kalman filtering and convex combination theory according to an embodiment of the present invention. DETAILED DESCRIPTION
[0067] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the relevant invention, rather than to limit the invention. It is also necessary to explain that, for ease of description, only the parts related to the relevant invention are shown in the accompanying drawings.
[0068] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.
[0069] The track fusion algorithm based on particle filter simulates a group of particles representing the possible states of the target, updates the weights of these particles according to the sensor observation data, and realizes accurate estimation of the target track. It can effectively handle nonlinear and non-Gaussian systems, adapt to multi-sensor data fusion, and handle model uncertainty. The track fusion algorithm based on joint probabilistic data association (JPDA) calculates the joint probability between multiple sensor observations and multiple target trajectories to achieve accurate fusion of tracks. It can effectively handle data association problems in multi-target environments and improve the accuracy and robustness of track tracking.
[0070] Kalman filtering is an algorithm that uses the linear system state equation and the system input and output observation data to optimally estimate the system state. Kalman filtering uses the state equation and observation equation, combined with the system's dynamic model and observation data, to optimally estimate the system state in a recursive manner.
[0071] The main steps are as follows: 1) predict the current state; 2) predict the covariance of the current state; 3) calculate the Kalman gain; 4) update the state estimate; 5) update the state covariance.
[0072] The convex combination fusion algorithm is a data fusion method. Its basic idea is to fuse the results of multiple data sources by weighted average to obtain more accurate results. The core of this algorithm lies in the convex combination theory, that is, for several weight coefficients, the sum of the above coefficients must be 1 and all must be non-negative.
[0073] The main steps are as follows: 1) Determine the data source and its weight: First, it is necessary to clarify the data sources involved in data fusion and their corresponding weights. The weight can be determined based on factors such as the reliability, coverage, and accuracy of the data source. 2) Data normalization: To facilitate subsequent calculations, the data from different data sources need to be normalized and converted into the same data range. 3) Convex combination calculation: Then, the normalized values of each data source are weighted averaged according to their weights to obtain a convex combination value. This convex combination value is the weighted average of the results of each data source, where the weight of each data source reflects its importance in the fusion result. 4) Denormalization: Finally, the obtained convex combination value is denormalized, that is, it is converted into the original data range to obtain the final fusion result.
[0074] Particle filters require a large number of particles to maintain the accuracy of estimation, which may lead to increased computational complexity and memory requirements. At the same time, particle degradation problems may lead to a decrease in particle diversity, affecting algorithm performance.
[0075] The computational complexity of the JPDA algorithm is high, especially in high target density environments, and the processing speed may be limited. At the same time, the dependence on prior probability and noise model may lead to performance degradation, especially in the case of high model uncertainty.
[0076] The Kalman filter is highly dependent on the accuracy of the system dynamic model and the observation noise model. If the model estimation is inaccurate, the performance of the filter may be affected. In addition, this method is sensitive to model changes and abnormal data, and its robustness is relatively low.
[0077] The convex combination fusion algorithm needs to perform weighted averaging of the outputs of different algorithms, which may require additional calculations and parameter adjustments. At the same time, the selection and adjustment of weights may depend on experience, which may affect the performance fluctuations of the algorithm.
[0078] In view of the shortcomings of the above-mentioned prior art, the embodiments of the present invention are specifically disclosed as follows:
[0079] First, Figure 1 It is shown that the embodiment of the present invention discloses a track fusion method based on Kalman filtering and convex combination theory, such as Figure 1 As shown, the method comprises the following steps:
[0080] S1. Data collection: collecting ship AIS data information collected by a base station in a certain place;
[0081] S2, data processing, data cleaning, data conversion and data aggregation processing of the collected ship AIS data information;
[0082] S3, track data generation, assuming that the ship's trajectory is a uniform linear motion, the track data is generated according to the object motion expression, in which the loss caused by external influences is considered to obtain the observation model;
[0083] S4, Kalman filter processing, using the Kalman filter algorithm to process the generated track data, the Kalman filter algorithm is composed of a time update equation and a measurement update equation, and the track information of each sensor is processed in turn according to a predetermined order until the prediction step of the track information of all sensors is completed;
[0084] S5, convex combination fusion, input the predicted trajectory information of each sensor at the end of the Kalman filter iteration, and use the convex combination algorithm to perform track fusion.
[0085] Specifically, the above steps are described in detail with reference to the accompanying drawings:
[0086] Step 1: Data processing
[0087] The data set used in this embodiment mainly includes the ship AIS information collected by several islands with base stations near Xia* Island (mainly Dayu, Houyu, Baituyu, Datuyu, and Xiaotuyu). The track AIS information mainly includes the latitude and longitude, speed, heading, time, and heading of the ship within the base station range. The collected track AIS data set is as follows: Figure 2 shown.
[0088] Data processing includes data cleaning, data conversion, and data aggregation.
[0089] (1) Data cleaning
[0090] Clean invalid, missing or erroneous data contained in AIS data, delete incomplete data, delete useless data, etc.
[0091] (2) Data conversion
[0092] Convert the information contained in AIS data into a data format that can be used for analysis. For example, convert a timestamp into a date and time format.
[0093] (3) Data Aggregation
[0094] The cleaned latitude and longitude location information of each ship is aggregated into higher-level data to form ship track information, and the ship track is marked on the GIS chart. Figure 3 shown.
[0095] Step 2: Algorithm training
[0096] (1) Generate track data
[0097] Assuming that the ship's trajectory is uniform linear motion, the object's motion expression is s = s0 + vt, where s represents the object's position at a certain moment, s0 is the object's initial position, v represents the object's speed, and t is time;
[0098] Expressed in matrix form: x = [0 t][x0 x] T , x is a vector representation related to the state of motion of the object, x0 is the initial position of the object; it is expressed as a state variable: x = [0 t][x k-1 x k-1 ] T =F k-1 x k-1 , x k-1 represents the state variable of the object at the k-1th moment, F k-1 is the state transfer matrix, which describes the transfer law of the object state from the k-1th moment to the next moment;
[0099] The loss caused by external influence is: xk =[0 t][x k-1 x k-1 ] T +[0 t]w k =F k-1 x k-1 +G k-1 w k-1 , x k represents the state variable of the object at the Kth moment, w k Represents external influencing factors, G k-1 is a matrix related to external influences, which describes the external influence factors w k-1 How to act on the state variables of an object;
[0100] The observation model is obtained as: C(k) is the observation model, which describes how to obtain the observation value from the actual state of the object, taking into account the state of external influences.
[0101] (2) Kalman filter algorithm
[0102] 1) Algorithm Overview
[0103] The Kalman filter algorithm consists of a time update equation and a measurement update equation. The time update equation can be understood as a prediction equation, and the measurement update equation can be understood as a correction equation.
[0104] In this embodiment, a sensor-to-system track fusion processing structure is used. This structure mainly involves data interaction and fusion processing between multiple sensors and a central processing system. Sensor-to-system track fusion simplifies the association problem into a bi-partite allocation problem, so common allocation algorithms can be used. However, the algorithm must deal with the problem of correlation estimation errors. Figure 4 In the example, the sensor track at point A and the system track at point B have correlated errors because both tracks depend on point C.
[0105] Furthermore, any error in the system track due to past processing errors in association or fusion will affect future fusion performance. Kalman filtering is used in the track association part. Kalman filtering is an algorithm that uses linear system state equations to make an optimal estimate of the system state through system input and output observation data. The time update equation is responsible for extrapolating the value of the current state variable and error covariance estimate forward in time to construct a priori estimates for the next time state. The measurement update equation is responsible for feedback, combining the prior estimate with the new measurement variable to construct an improved prior estimate.
[0106] Through the state equation and observation equation, combined with the system's dynamic model and observation data, the state of the system is optimally estimated using a recursive method. The Kalman filter process can be divided into two steps: prediction and update. In the prediction step, the state value at the current moment is predicted based on the state estimate value and state transfer matrix at the previous moment; in the update step, the predicted state value is corrected based on the observation value and observation matrix at the current moment to obtain the state estimate value at the current moment. At the same time, the estimation result is continuously optimized by minimizing the covariance of the estimation error, such as Figure 5 As shown in the state equation, x k represents the state quantity at the kth moment, Z k Represents the observation value at time k, and matrix H is the observation matrix.
[0107] Specifically, the time update equation is responsible for calculating the value of the current state variable and the error covariance estimate forward in time, and constructing a priori estimates for the next time state. The measurement update equation is responsible for feedback, combining the prior estimates with the new measurement variables to construct improved prior estimates.
[0108] ① Time update equation:
[0109]
[0110] in, represents the predicted value of the system state at time k based on the observation information at time k-1 and before, F k-1 is the state transfer matrix, which describes the transition law of the system state from time k-1 to time k. is the estimated value of the system state at time k based on the observation information at time k-1, P k|k-1 G represents the covariance matrix of the system state prediction value at time k based on the observation information at time k-1 and before, k-1 The matrix related to system noise describes how system noise affects the transition of system state, Q k-1 is the covariance matrix of the system noise, which is used to describe the statistical characteristics of the system noise.
[0111] ②Measurement update equation:
[0112]
[0113] Among them, K k is the Kalman gain matrix, which is used to weigh the predicted value and the observed value in the measurement update process. k is the observation matrix, which describes the linear relationship between the system state and the observation value, Z k is the observed value at time k, R k is the covariance matrix of the observation noise, which is used to describe the statistical characteristics of the observation noise. represents the updated estimate of the system state at time k based on the observation information at time k, P k|k Represents the covariance matrix of the updated estimate of the system state at time k based on the observation information at time k.
[0114] 2) Algorithm Process
[0115] ① Input pre-processed data (ship’s coordinate information).
[0116] ② Define the initial values of the system's state variables (such as position, velocity, etc.) and observation data. Set the initial state estimate and covariance matrix.
[0117]
[0118] In the formula, represents the estimated value of the system state based on the initial information at the initial time (k=0), E[x(0)] is the mathematical expectation of the system at the initial time x(0), p(0|0) represents the covariance matrix of the estimated value of the system state at the initial time (k=0), and x(0) is the actual state of the system at the initial time.
[0119] ③Use the system's state transition equation to predict the state at the next moment. Update the covariance matrix of the predicted state.
[0120] ④ Use the covariance matrix of the predicted state and the covariance matrix of the observation noise to calculate the Kalman gain.
[0121] ⑤ Use the Kalman gain and observations to update the state estimate and the covariance matrix of the state estimate. Use the updated state estimate and covariance matrix as the initial values for the next iteration and repeat the prediction and update steps.
[0122] ⑥ Determine whether the end condition is met, that is, whether the track information prediction of all sensors has been completed.
[0123] Step 3: Track Fusion
[0124] (1) Convex combination algorithm formula:
[0125]
[0126] in, represents the estimated value of the system state after fusion, N is the number of sensors, P i is the state estimation covariance matrix corresponding to the i-th sensor, (P i ) -1 is the inverse matrix of the i-th sensor state estimation covariance matrix, is the estimated value of the system state by the i-th sensor; P -1is the inverse matrix of the system state estimation covariance matrix after fusion, and P is the system state estimation covariance matrix after fusion.
[0127] (2) Convex combination algorithm process:
[0128] ① Input the predicted trajectory information of each sensor at the end of the Kalman filter iteration.
[0129] ②Initialize the sensor weights.
[0130] ③Based on the set weights, perform weighted summation on each predicted trajectory. Multiple inputs are fused into one output through linear combination.
[0131] ④ Based on the evaluation results, the weights, fusion strategies and other parameters can be adjusted to optimize the fusion effect.
[0132] If the operation results are as follows Figure 6-8 As shown, Figure 6 (a) and Figure 6 (b) X-coordinate position tracking error and Y-coordinate position tracking error respectively; Figure 7 is a schematic diagram of position error; Figure 8 is the aggregated ship trajectory, such as Figure 8 shown.
[0133] The embodiment of the present invention adopts a sensor-to-system track fusion processing structure, sends all data to a central system for unified fusion processing, simplifies the association problem, and makes it a bi-partite allocation problem. Common allocation algorithms can be used to avoid the inefficiency problem that may be caused by discarding past processing results in the sensor-to-sensor track fusion structure.
[0134] The use of Kalman filtering for track association has relatively loose requirements on the statistical properties of system disturbances and observation errors, and does not require the assumption that the signal and noise must be a stationary process. At the same time, it can process the observed signal containing noise to obtain an estimate of the real signal with the minimum error, thereby improving the accuracy of track association.
[0135] A convex combination fusion algorithm is used for track state estimation. Based on the convex combination theory, the results of multiple data sources are fused by weighted average to make the results more accurate. At the same time, the algorithm clarifies the method of determining the data source and its weight, as well as the specific steps of data normalization, convex combination calculation and denormalization, and has stronger operability and accuracy.
[0136] In summary, the technology of the present invention is technologically advanced in terms of processing structure, track association and track state estimation fusion, and can improve the effect and accuracy of track fusion.
[0137] Further references Fig. 9 As an implementation of the methods shown in the above figures, the present application provides an embodiment of a device, and the device embodiment is Figure 1 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.
[0138] In a second aspect, an embodiment of the present invention provides a track fusion system based on Kalman filtering and convex combination theory, such as Fig. 9 As shown, it includes: a data acquisition module 91, a data processing module 92 and a track data generation module 93.
[0139] In a specific embodiment, the data acquisition module 91 is configured to collect ship AIS data information collected by a base station at a certain location; the data processing module 92 is configured to perform data cleaning, data conversion and data aggregation processing on the collected ship AIS data information; the track data generation module 93 is configured to assume that the ship trajectory is a uniform linear motion, and generate track data based on the object motion expression, wherein the loss caused by external influences is taken into account to obtain an observation model.
[0140] The functions of the above modules correspond to the methods and will not be repeated here.
[0141] As another aspect, the present invention also provides a computer-readable medium, which may be included in the electronic device described in the above embodiment; or it may exist independently and not be assembled into the electronic device. The above computer-readable medium carries one or more programs. When the above one or more programs are executed by the electronic device, the electronic device executes the method steps described in the first aspect, including: data acquisition, collecting ship AIS data information collected by a base station in a certain place; data processing, data cleaning, data conversion and data aggregation processing of the collected ship AIS data information; track data generation, assuming that the ship trajectory is a uniform linear motion, and generating track data according to the object motion expression, wherein the loss caused by external influence is considered to obtain the observation model; Kalman filter processing, using the Kalman filter algorithm to process the generated track data, the Kalman filter algorithm consists of a time update equation and a measurement update equation, and processes the track information of each sensor in turn in a predetermined order until the prediction step of the track information of all sensors is completed; convex combination fusion, inputting the predicted track information of each sensor at the end of the Kalman filter iteration, and using the convex combination algorithm for track fusion.
[0142] The above description is only a preferred embodiment of the present invention and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the present invention is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above inventive concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present invention (but not limited to) to form a technical solution.
Claims
1. A track fusion method based on Kalman filtering and convex combination theory, characterized in that: The method comprises the following steps: Data collection: collect ship AIS data information collected by a base station in a certain place; Data processing, data cleaning, data conversion and data aggregation processing of the collected ship AIS data information; Track data generation: Assuming that the ship's trajectory is a uniform linear motion, the track data is generated based on the object motion expression, taking into account the loss caused by external influences to obtain the observation model; Kalman filter processing, using the Kalman filter algorithm to process the generated track data, the Kalman filter algorithm consists of a time update equation and a measurement update equation, and processes the track information of each sensor in turn according to a predetermined order until the prediction step of the track information of all sensors is completed; Convex combination fusion, input the predicted trajectory information of each sensor at the end of the Kalman filter iteration, and use the convex combination algorithm to perform track fusion.
2. The track fusion method based on Kalman filtering and convex combination theory according to claim 1, characterized in that: The track data generation specifically includes: Assuming that the ship's trajectory is uniform linear motion, the object's motion expression is s = s0 + vt, where s represents the object's position at a certain moment, s0 is the object's initial position, v represents the object's speed, and t is time; Expressed in matrix form: x = [0 t][x0 x] T , x is a vector representation related to the state of motion of the object, x0 is the initial position of the object; it is expressed as a state variable: x = [0 t][x k-1 x k-1 ] T =F k-1 x k-1 , x k-1 represents the state variable of the object at the k-1th moment, F k-1 is the state transfer matrix, which describes the transfer law of the object state from the k-1th moment to the next moment; The loss caused by external influence is: x k =[0 t][x k-1 x k-1 ] T +[0 t]w k =F k-1 x k-1 +G k-1 w k-1 , x k represents the state variable of the object at the Kth moment, w k Represents external influencing factors, G k-1 is a matrix related to external influences, which describes the external influence factors w k-1 How to act on the state variables of an object; The observation model is obtained as: C(k) is the observation model, which describes how to obtain the observation value from the actual state of the object, taking into account the state of external influences.
3. The track fusion method based on Kalman filtering and convex combination theory according to claim 1, characterized in that: The Kalman filter algorithm consists of a time update equation and a measurement update equation. The time update equation is: in, represents the predicted value of the system state at time k based on the observation information at time k-1 and before, F k-1 is the state transfer matrix, which describes the transition law of the system state from time k-1 to time k. is the estimated value of the system state at time k based on the observation information at time k-1, P k|k-1 G represents the covariance matrix of the system state prediction value at time k based on the observation information at time k-1 and before, k-1 The matrix related to system noise describes how system noise affects the transition of system state, Q k-1 is the covariance matrix of the system noise, which is used to describe the statistical characteristics of the system noise; The measurement update equation is: Among them, K k is the Kalman gain matrix, which is used to weigh the predicted value and the observed value in the measurement update process. k is the observation matrix, which describes the linear relationship between the system state and the observation value, Z k is the observed value at time k, R k is the covariance matrix of the observation noise, which is used to describe the statistical characteristics of the observation noise. represents the updated estimate of the system state at time k based on the observation information at time k, P k|k Represents the covariance matrix of the updated estimate of the system state at time k based on the observation information at time k.
4. The track fusion method based on Kalman filtering and convex combination theory according to claim 3 is characterized in that: The Kalman filter processing uses the Kalman filter algorithm to process the generated track data. The Kalman filter algorithm consists of a time update equation and a measurement update equation, and sequentially completes the prediction of the track information of all sensors, specifically including: Input the pre-processed data, i.e. the coordinate information of the ship; Define the initial values of the system's state variables such as position, velocity, and observation data, and set the initial state estimates and covariance matrix: In the formula represents the estimated value of the system state based on the initial information at the initial time (k = 0), E[x(0)] is the mathematical expectation of the system at the initial time x(0), p(0|0) represents the covariance matrix of the estimated value of the system state at the initial time (k = 0), and x(0) is the actual state of the system at the initial time; Use the system's state transition equation to predict the state at the next moment and update the covariance matrix of the predicted state; The Kalman gain is calculated using the covariance matrix of the predicted state and the covariance matrix of the observation noise; Use the Kalman gain and the observation value to update the state estimate and the covariance matrix of the state estimate, use the updated state estimate and covariance matrix as the initial value of the next iteration, and repeat the prediction and update steps; Determine whether the end condition is met, that is, whether the track information prediction of all sensors is completed.
5. The track fusion method based on Kalman filtering and convex combination theory according to claim 1, characterized in that: The algorithm formula of the convex combination fusion is: in, represents the estimated value of the system state after fusion, N is the number of sensors, P i is the state estimation covariance matrix corresponding to the i-th sensor, (P i ) -1 is the inverse matrix of the i-th sensor state estimation covariance matrix, is the estimated value of the system state by the i-th sensor; P -1 is the inverse matrix of the system state estimation covariance matrix after fusion, and P is the system state estimation covariance matrix after fusion.
6. The track fusion method based on Kalman filtering and convex combination theory according to claim 5, characterized in that: The convex combination fusion inputs the predicted trajectory information of each sensor after the Kalman filter iteration, and uses the convex combination algorithm to perform track fusion, specifically including: Input the predicted trajectory information of each sensor at the end of the Kalman filter iteration; Initialize the sensor weights; Based on the set weights, each predicted trajectory is weighted summed, and multiple inputs are fused into one output through linear combination; Based on the evaluation results, feedback is provided to adjust the weights and parameters of the fusion strategy to optimize the fusion effect.
7. The track fusion method based on Kalman filtering and convex combination theory according to claim 1, characterized in that: The data processing includes data cleaning, data conversion and data aggregation processing on the collected ship AIS data information, specifically including: Data cleaning: clean the collected ship AIS data, use preset rules and algorithms to accurately identify and remove invalid, missing or erroneous data, and delete incomplete and useless data; Data conversion: converting the information contained in the cleaned AIS data into a data format that can be used for analysis, such as converting a timestamp into a specific date and time format; Data aggregation: Aggregate the converted data and use a specific algorithm to aggregate the longitude and latitude position information of each ship into ship track information, and mark the ship track on the GIS chart.
8. A track fusion system based on Kalman filtering and convex combination theory, characterized in that: include: A data collection module is configured to collect ship AIS data information collected by a base station at a certain location; A data processing module is configured to perform data cleaning, data conversion and data aggregation processing on the collected ship AIS data information; The track data generation module is configured to assume that the ship's trajectory is a uniform linear motion, and to generate track data based on the object motion expression, wherein the loss caused by external influences is taken into account to obtain an observation model.
9. An electronic device, comprising: one or more processors; A storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.