A Distributed 3D Extended Target Tracking Method Based on Feature Selection

Abstract

This invention discloses a distributed 3D extended target tracking method based on feature selection, comprising the following steps: acquiring the raw measurements of each sensor at time k; preprocessing the raw measurements to obtain preprocessed raw measurements; inputting the preprocessed raw measurements into the GIW-PHD filtering algorithm to obtain the estimated state information of the target detected by each sensor; preprocessing the estimated state information of the detected target to obtain the target state of each sensor; and selecting the target state for feature extraction and fusion to obtain the target state information at time k. This invention can shorten the data filtering operation time while performing feature selection and fusion of the target state of the sensors, and improve the estimation accuracy of the position information and 3D extended information of the 3D extended target by using feature splicing and fusion.

Description

Technical Field

[0001] This invention belongs to the field of target tracking technology, specifically relating to a distributed three-dimensional extended target tracking method based on feature selection. Background Technology

[0002] 3D target tracking is widely used in fields such as autonomous driving and robot navigation. Common 3D target tracking methods include vision-based methods, model-based methods, and multi-sensor-based methods. As sensor resolution increases, the shape and size of 3D targets cannot be ignored, making them 3D extended targets with size and shape.

[0003] Multi-sensor-based 3D extended target tracking methods include two types: centralized tracking methods and distributed methods. Centralized tracking methods, such as the paper "Extended Target GIW-PHD Tracking Algorithm Based on Multi-Sensor Information Fusion," fuse measurement data from multiple sensors to ensure the accuracy of state information estimation for the 3D extended target. However, due to the large amount of measurement data for the 3D extended target in centralized fusion algorithms, the transmitted data is large, resulting in weak robustness and scalability.

[0004] Distributed tracking methods, such as the patent document "A Distributed Multi-Sensor Fusion Tracking Method and System" with publication number CN116028890A, obtain the probability posterior density through a local extended label multi-Bernoulli filter, fuse the local posterior probability densities of multiple sensors based on the relative weight matrix, and update the target state after obtaining the fused probability posterior density to finally determine the distributed multi-sensor fusion tracking result. This method performs filtering before fusion, which reduces the size of transmitted data and improves the robustness and scalability of the network by eliminating the need for a specific fusion center. However, it does not take into account that the illumination surface of the target is different under multiple sensors, and the acquired measurement data is different. Therefore, this method is not accurate enough in estimating the state of the three-dimensional extended target. Summary of the Invention

[0005] To address the aforementioned problems in the existing technology, this invention proposes a distributed three-dimensional extended target tracking method based on feature selection.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solution:

[0007] This invention relates to a distributed 3D extended target tracking method based on feature selection, comprising the following steps:

[0008] Measurement data acquisition steps: Acquire the raw measurement data of each sensor that detects the target at time k;

[0009] Measurement data preprocessing steps: Preprocess the raw measurement data of each sensor to obtain the preprocessed raw measurement data of each sensor.

[0010] Steps for obtaining estimated state information: Input the preprocessed raw measurement data of each sensor into the GIW-PHD filtering algorithm to obtain the estimated state information of the target detected by each sensor;

[0011] Estimated state information preprocessing steps: The estimated state information of the target detected by each sensor is preprocessed to obtain the target state of each sensor;

[0012] Target state information acquisition steps: Select the target state of each sensor, extract and fuse features to obtain the fused target state information at time k.

[0013] Furthermore, the raw measurement data of each sensor consists of the target measurement source and Gaussian white noise.

[0014] Furthermore, the measurement data preprocessing steps include:

[0015] The original measurement data of each sensor are initialized separately to obtain the original measurement data of each sensor after initialization;

[0016] The original measurement data of each sensor after initialization are projected onto the qz plane corresponding to each sensor to obtain the original measurement data of each sensor after preprocessing.

[0017] Furthermore, the measurement data preprocessing step also includes:

[0018] The initial measurement data of each sensor after initialization are projected onto the xy plane to obtain the initial measurement projection of each sensor.

[0019] The mean value of the initial measurement projection of each sensor and the yaw angle of each sensor when detecting the target are determined by using the initial measurement projection of each sensor.

[0020] Based on the mean value of the initial measurement projection of each sensor and the yaw angle of each sensor when detecting the target, determine the straight line q corresponding to each sensor. Based on the straight line q and z-axis corresponding to each sensor, determine the qz plane corresponding to each sensor.

[0021] Furthermore, the yaw angle of each sensor when detecting the target is determined using the initial measurement projection of each sensor, including:

[0022] The initial measurement projections of each sensor are centered to obtain the centered initial measurement projections of each sensor.

[0023] Calculate the covariance matrix of the initial measurement projection after centering each sensor. The angle between the eigenvector corresponding to the covariance matrix and the x-axis is the yaw angle of each sensor when detecting the target.

[0024] Furthermore, the steps for estimating state information acquisition include:

[0025] The target's state and covariance at time k-1 are substituted into the GIW-PHD filtering algorithm to predict the predicted GIW components.

[0026] The original measurement data after preprocessing of each sensor are fed into the GIW-PHD filtering algorithm to update the predicted GIW components, thus obtaining the updated GIW components for each sensor.

[0027] Based on the set trimming threshold, the merging threshold between components, and the maximum number of GIW components, the updated GIW components of each sensor are trimmed and merged to obtain the new GIW components of each sensor.

[0028] The estimated state information of the target detected by each sensor is obtained by selecting the GIW components with a weight greater than 0.5 from the new GIW components of each sensor.

[0029] Furthermore, the estimated state information preprocessing steps include:

[0030] Based on the estimated state information of the target detected by each sensor, the ellipse estimation state of the extended target corresponding to each sensor is obtained respectively;

[0031] The target state of each sensor is obtained by combining the elliptical estimated state of the extended target corresponding to each sensor with the yaw angle when each sensor detects the target.

[0032] Furthermore, the steps for obtaining target state information include:

[0033] Feature extraction and fusion are performed on the target states of the two sensors with the largest difference in major axis among the target states of each sensor to obtain the extended state information of the fused target;

[0034] By using the extended state information of the fused target, the target center position is fused to obtain the fused target state information at time k.

[0035] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0036] 1. This invention estimates the yaw angle by acquiring measurement data from various sensors used to detect the target, and simultaneously centers the measurement data. A new two-dimensional coordinate system is constructed using the yaw angle, the centered measurement data, and the z-axis. The measurement data is then projected onto this two-dimensional coordinate system. By reducing the dimensionality of the original sensor measurement data, the processing time of the data in GIW-PHD filtering is greatly shortened.

[0037] 2. This invention obtains the estimated state information of the target detected by the sensor through filtering and screening. Combined with the yaw angle of each sensor when detecting the target, the target state of each sensor is obtained. The target states of the two sets of sensors with the largest difference in the major axis are selected for feature selection and fusion. By using feature splicing and fusion, the estimation accuracy of the position information and the three-dimensional extended information of the target is improved.

[0038] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0039] Figure 1 This is an overall flowchart of the method of the present invention.

[0040] Figure 2 This is a comparison diagram of the target center position tracking results in the x, y, and z directions using the method of this invention, and the actual target trajectory.

[0041] Figure 3 This is a comparison chart of the root mean square error of the target position obtained by the method of this invention and other radars after 100 Tecalo tests.

[0042] Figure 4 This is a comparison chart of the root mean square error of the method of this invention and the target spread state obtained by other radars after 100 Tecalo tests.

[0043] Figure 5 This is a comparison diagram of the target expansion state estimated by the method of the present invention and the actual target expansion state at a certain moment. Detailed Implementation

[0044] To further illustrate the technical means and effects adopted by the present invention to achieve the intended purpose, the following detailed description of the solution according to the present invention is provided in conjunction with the accompanying drawings and specific embodiments.

[0045] The foregoing and other technical contents, features, and effects of the present invention will be clearly presented in the following detailed description of specific embodiments in conjunction with the accompanying drawings. Through the description of the specific embodiments, a more in-depth and concrete understanding can be gained of the technical means and effects adopted by the present invention to achieve its intended purpose. However, the accompanying drawings are for reference and illustration only and are not intended to limit the technical solutions of the present invention.

[0046] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations are intended to cover non-exclusive inclusion, such that an article or apparatus comprising a list of elements includes not only those elements but also other elements not expressly listed. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the article or apparatus that includes said element.

[0047] like Figure 1 As shown, this embodiment provides a distributed 3D extended target tracking method based on feature selection, which specifically includes the following steps:

[0048] Step 1: Obtain the original measurement data of each sensor that detects the target at time k. The original measurement data of the sensor consists of the target measurement source and Gaussian white noise.

[0049] The raw measurements of each sensor on the target at time k are as follows:

[0050] z k =ζ k +v

[0051] Among them, z k These are the raw measurement data from the sensor, where v is zero-mean Gaussian white noise, and ζ is... k Let ζ be the target measurement source at time k. The target measurement source ζ can be represented as:

[0052] ζ=s·A·U1·U2·U3

[0053] A=[a·sinβcosγ,b·sinβsinγ,c·cosβ]

[0054]

[0055] Where A is the standard parametric equation vector of the ellipsoid, and β∈(0,π) and γ∈(0,2π); U1, U2, and U3 are the rotation moments corresponding to the x, y, and z axes, respectively; w1 is the roll angle, w2 is the pitch angle, w3 is the yaw angle, a is the major axis, b is the minor axis, and c is the median axis; s∈(0,1) is the scaling factor; β, γ, and s are random variables.

[0056] Step 2: Preprocess the original measurement data of the sensor to obtain the original measurement data of the sensor after preprocessing.

[0057] Step 2.1: Initialize the raw measurements of each sensor to obtain the initialized raw measurement data of each sensor. Since each sensor detects only one side of the target, it only contains a portion of the measurements in the target measurement model. Taking the l-th sensor as an example, the initialized raw measurement data detected by the l-th sensor at time k is as follows:

[0058]

[0059] in, Let x represent the i-th measurement generated by the l-th sensor. i The x-coordinate of the i-th measurement generated by the l-th sensor is represented by the y-coordinate. i This shows the coordinates of the i-th measurement generated by the l-th sensor on the y-axis, z. i Let i represent the coordinate of the i-th measurement generated by the l-th sensor on the z-axis, where i is the number of measurements generated by the l-th sensor at time k.

[0060] Step 2.2: Project the initial measurement data of the l-th sensor onto the xy plane to obtain the initial measurement projection of the l-th sensor. Specifically, the initial measurement data... In The measurement projected onto the xy plane is represented as:

[0061]

[0062] in, This is the initial measurement projection of the l-th sensor. The i-th measurement generated by the l-th sensor projected onto the xy-axis plane.

[0063] Step 2.3: Use the initial measurement projection of the l-th sensor to determine the mean value of the initial measurement projection of the l-th sensor and the yaw angle when the l-th sensor detects the target.

[0064] Step 2.3.1: Center the initial measurement projection of the l-th sensor to obtain the centered initial measurement projection of the l-th sensor:

[0065]

[0066] in, The initial measurement projection of the l-th sensor The mean.

[0067] Step 2.3.2: Calculate the covariance matrix of the initial measurement projection after centering the l-th sensor. covariance matrix The corresponding feature vector is r z,max A 2×1 dimensional vector rz,max The angle between the x-axis and the x-axis is the yaw angle when the l-th sensor detects the target.

[0068] Step 2.4: Determine the straight line q based on the mean value of the initial measurement projection of the l-th sensor and the yaw angle of the l-th sensor when detecting the target. The straight line q and the z-axis determine the qz plane. Project the original measurement data of the l-th sensor after initialization onto the qz plane to obtain the original measurement data of the l-th sensor after preprocessing. Specifically:

[0069] Passing point The angle with the x-axis is Construct a new plane qz with the straight line q and the z-axis, and measure... In If the measurement is projected onto the plane, then the original measurement data after preprocessing by the l-th sensor can be expressed as:

[0070]

[0071] Where, q i This represents the coordinates of the original measurement data on the q-axis after preprocessing, z. i This represents the coordinates of the original measurement data on the z-axis after preprocessing. The formula for calculating the parameter θ is: Where, x i The x-coordinate of the i-th measurement generated by the l-th sensor is represented by the y-coordinate. i This shows the coordinates of the i-th measurement generated by the l-th sensor on the y-axis.

[0072] Step 3: Input the preprocessed raw measurement data into the GIW-PHD filtering algorithm to obtain the estimated state information of the target detected by the sensor.

[0073] Specifically, the original measurement data after preprocessing of the l-th sensor in step 2, i.e., the measurement at time k in the l-th sensor. The estimated state information of the target detected by the l-th sensor is obtained by substituting it into the GIW-PHD filtering algorithm.

[0074] Step 3.1: Substitute the target's state and covariance at time k-1 into the GIW-PHD filtering algorithm to predict the predicted GIW component.

[0075] Based on the target state and covariance at time k-1, the prediction of PHD can be expressed as:

[0076] D k|k-1 (ξ k ) = D S,k|k-1 (ξ k-1 )+γk (ξ k )

[0077] Among them, D S,k|k-1 (ξ k-1 ) represents the predicted GIW-PHD of the target surviving at time k, γ k (ξ k ) represents the GIW-PHD of the newborn target at time k, ξ k =(x k ,X k ) represents the state of the target, x k Let X be the motion state vector of the target's centroid. k The extended state of the target.

[0078] The predicted GIW-PHD for survival targets is:

[0079]

[0080] Where, p S,k Indicates the probability of the target's survival. Let N(x; m, P) represent the weight component of the i-th GIW component at time k-1, N(x; m, P) represent the mean m, and P is a Gaussian distribution with covariance matrix P. Let IW(X; v, V) represent an inverse Wieschat distribution with degrees of freedom v and a scale matrix V. The parameters in the formula are expressed as follows:

[0081]

[0082] Among them, F k-1 Let Q be the state transition matrix. k-1 For process noise, T s The sampling time is represented by τ, the time decay factor is τ, and the target expansion dimension is d. These represent the motion state, covariance matrix, inverse Wieschat degrees of freedom, and inverse Wieschat scale matrix of the surviving target.

[0083] The predicted GIW–PHD for the newborn target set is:

[0084]

[0085] in, These represent the motion state, covariance matrix, inverse Wieshardt degrees of freedom, and inverse Wieshardt scale matrix of the new target.

[0086] Step 3.2: Input the original measurement data of the l-th sensor after preprocessing in Step 2 into the GIW-PHD filtering algorithm to update the predicted GIW components, and obtain the updated GIW components of the l-th sensor. The updated GIW-PHD is then expressed as:

[0087]

[0088] in, This indicates the missed detection portion of GIW-PHD. It uses a subset of measurements W to update ξ. k GIW-PHD after that, p∠Z k This indicates that p will measure the set Z. k Divide into a non-empty unit set W; It can be represented as:

[0089]

[0090] in, ξ k The update formula is expressed as:

[0091]

[0092] ξ k =ξ k|k-1

[0093] in, ξ k Let p represent the weight and parameter of the j-th GIW component, respectively, where γ(j) is the measurement rate of the j-th component. D It is the probability of detecting the measurement.

[0094] It can be represented as:

[0095]

[0096] In the formula, the update parameter is expressed as:

[0097]

[0098] in, These are the motion state, covariance matrix, and inverse Wissaud scale matrix of the j-th GIW component using the measurement subset W and the newly completed GIW component, respectively. Let W be the mean and measurement covariance matrix of the measurement subset W, respectively. These are the innovation factor, update gain matrix, central measurement innovation, and innovation matrix obtained by updating the j-th GIW component using the measurement subset W, respectively.

[0099] The likelihood function is:

[0100]

[0101] Where |V| represents the determinant of the matrix, and |W| represents the number of measurements in the measurement subset W. This is the likelihood value of the j-th GIW component after being updated using the measurement subset W.

[0102] The updated weights are:

[0103]

[0104] Where, β FA,k It is the clutter number generated in the unit survival space of the scan. This represents the j-th weight updated in the measurement subset W. p is the likelihood value of the j-th GIW component after the measurement subset W is updated. D Indicates the probability that the target is detected:

[0105]

[0106] Where, δ i,j For the Kronecker delta function, w p This is represented as the partition weight.

[0107] Step 3.3: Based on the set trimming threshold Tr, the component merging threshold U, and the maximum number of GIW components J. max The updated GIW components of the l-th sensor are trimmed and merged to obtain the new GIW components of the l-th sensor;

[0108] Specifically, the trimming threshold is set to Tr, the merging threshold between components is U, and the maximum number of GIW components is J. max ; Discard GIW components with weights less than the pruning threshold Tr; merge GIW components whose Euclidean distance between them is less than U; if the final number of GIW components is greater than J. max Select the final GIW component J max The term with the largest weight is pruned and merged to obtain the new GIW component of the l-th sensor after merging.

[0109] Step 3.4: Select the weights in the new GIW component of the l-th sensor. The GIW component is used to obtain the estimated state information of the target detected by the l-th sensor. Represented as:

[0110]

[0111] in, Let J represent the motion state estimate and extended information estimate of the j-th target at time k, respectively, where J represents the total number of estimated targets. Let d be the inverse Wieshard degree of freedom of the j-th objective, and d be the objective extension dimension.

[0112] Step 4: Preprocess the estimated state information of the target detected by the l-th sensor to obtain the target state of the l-th sensor.

[0113] Step 4.1: Based on the estimated state information of the target detected by the l-th sensor. Obtain the elliptic estimated state of the extended target corresponding to the l-th sensor at time k.

[0114] Specifically, from formula X k =R k diag([a 2 ,b 2 ])R k T Where a and b are the major and minor axes, and R k Using the rotation matrix, we obtain the elliptical estimated state of the extended target at time k. in, It is the major axis of the ellipse shape of the j-th target. It is the minor axis of the ellipse shape of the j-th target. It is the angle between the j-th target and the coordinate axis q.

[0115] Step 4.2: Estimate the state of the target by extending the ellipse at time k. Yaw angle compared to when the l-th sensor detects the target in step 2 In this embodiment, taking the detection of the j-th target by the l-th sensor as an example, the target state obtained by the l-th sensor is... for:

[0116]

[0117] in, It is the major axis of the ellipse shape of the j-th target in the l-th sensor. It is the minor axis of the elliptical shape of the j-th target in the l-th sensor. It is the angle between the j-th target in the l-th sensor and the coordinate axis q. It is the yaw angle of the j-th target in the l-th sensor. It is the position information of the j-th target in the l-th sensor. It is the position information of the j-th target in the l-th sensor on the coordinate axis q. It is the position information of the j-th target in the l-th sensor on the z-axis, and this position information is in the qz plane.

[0118] From the formula Location information can be Transform into the xyz three-dimensional coordinate system, i.e. in, It is the position information of the j-th target in the l-th sensor on the x-axis. It is the position information of the j-th target in the l-th sensor on the y-axis. It is the position information of the j-th target in the l-th sensor on the z-axis.

[0119] Step 5: Select the target state of the sensor, extract and fuse features to obtain the target state information at time k.

[0120] Step 5.1: Simultaneously perform steps 1 to 4 above to obtain the target state of all sensors that detect the target. In this embodiment, each sensor obtains the state of the j-th target. Select the target states of the two sensors with the largest difference in major axis, sensor 1 and sensor 2, for feature extraction and fusion.

[0121] Sensor 1 is represented by the longer axis of the two sensors, and its target state is represented as follows:

[0122]

[0123] in, Let j be the major axis of the extended target j in sensor 1. Let the minor axis of the extended target j in sensor 1 be... To extend the angle between target j and the xy plane in sensor 1, To extend the yaw angle of target j in sensor 1, To extend the position information of target j in sensor 1.

[0124] Sensor 2 is represented by the shorter of the two sensors, and its target state is represented as follows:

[0125]

[0126] in, Let the major axis of the extended target j in sensor 2 be... For the minor axis of the extended target j in sensor 2, To extend the angle between target j and the xy plane in sensor 2, To extend the angle between the yz planes of target j in the two sensors, To extend the position information of target j in sensor 2.

[0127] Feature extraction is performed on sensors 1 and 2, and the major axis of the extended target j in sensor 1 is... Long axis as the target of integration and expansion The major axis of extended target j in sensor 2 The short axis as a target for fusion expansion Fusion of the minor axis of extended target j from the two sensors The central axis as the goal of integration Extend the target j in sensor 1 And the extended target j in sensor 2 Each is used as the pitch angle of the fusion target. and roll angle Extend the target j in sensor 1 The yaw angle of the fusion target is used to obtain the extended state information of the fusion target.

[0128] The formula for the extended state information of the fused target is expressed as follows:

[0129] Long axis short axis Central axis Yaw angle Pitch angle Roll angle

[0130] Step 5.2: Use the state extension information of the fused target to fuse the target center position and obtain the fused target state information at time k.

[0131] Specifically, the position information of target j in sensor 1 is extended. With a 90° yaw angle Construct a spatial line λ1 parallel to the xy plane, with the parametric equation of λ1 expressed as:

[0132]

[0133] The position information of extended target j in sensor 2 With a 90° yaw angle Construct a spatial line λ2 parallel to the xy plane, with the parametric equation of λ2 expressed as:

[0134]

[0135] The z-direction values ​​of the two spatial lines are merged into... Solving the parametric expressions of the two spatial lines λ1 and λ2 simultaneously, the fused target state information at time k is represented as:

[0136] m k ={x k ,y k ,z k}

[0137] Where, x k Let be the coordinates of the fused target position information on the x-axis at time k, and y be the coordinates of the target position information on the x-axis. kLet z be the coordinates of the fused target position information on the y-axis at time k, and z be the coordinates of the target position information on the z-axis. k Let k be the coordinates of the fused target location information on the z-axis.

[0138] This invention estimates yaw angles by acquiring sensor measurement data, centers the measurement data, and constructs a new two-dimensional coordinate system using the yaw angle, the centered measurement data, and the z-axis. Projecting the measurement data into this two-dimensional coordinate system reduces the data dimensionality and significantly shortens the processing time in GIW-PHD filtering. Simultaneously, GIW-PHD filtering is used to filter and obtain the estimated state information of the target detected by the sensor. The estimated state information of the target is combined with the yaw angle when the sensor detects the target to obtain the target state of the sensor. The target states of the two sets of sensors with the largest difference in the major axis are selected for feature selection and fusion. By using feature splicing and fusion, the estimation accuracy of the position information and the three-dimensional extended information of the target is improved.

[0139] Figure 2 This is a comparison diagram of the target center position tracking results in the x, y, and z directions using the method of this invention, and the actual target trajectory. The target moves at a constant speed in the three directions. The circled lines in the diagram represent the estimated results, and the solid lines represent the actual target trajectory. It can be seen that this method can accurately estimate the center position of the three-dimensional extended target.

[0140] Figure 3 This is a comparison chart of the root mean square error (RMSE) of the target position obtained by the method of this invention and other radars after 100 Tecalo trials. It shows the RMSE of the position obtained by filtering the two radars individually and the RMSE obtained by the distributed fusion of the method of this invention. It can be clearly seen that the method of this invention significantly improves the position estimation accuracy.

[0141] Figure 4 This is a comparison chart of the root mean square error (RMSE) of the target spread state obtained by the method of this invention and other radars after 100 Tecalo trials. It shows the RMSE of the spread state obtained by individual filtering of the two radars and the RMSE of the spread state obtained through distributed fusion using the method of this invention. It can be seen that the method of this invention significantly improves the accuracy of spread state estimation.

[0142] Figure 5 This is a comparison diagram between the target's spread state estimated by the method of this invention and the actual target spread state at a certain moment. At this moment, the actual target's yaw angle is 45°, roll angle is 0°, and pitch angle is 0°. It can be seen that the method of this invention can accurately estimate the target's spread state.

[0143] Finally, it should be noted that the above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention and should not be construed as limiting the specific implementation of the invention to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the inventive concept, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A feature selection based distributed three-dimensional extended target tracking method, characterized in that, The method comprises the following steps: a measurement data acquisition step: acquiring original measurement data of each sensor detecting the target at time k; a measurement data preprocessing step: preprocessing the original measurement data of each sensor to obtain preprocessed original measurement data of each sensor; initializing the original measurement data of each sensor to obtain initialized original measurement data of each sensor; projecting the initialized original measurement data of each sensor onto an x-y plane to obtain initial measurement projections of each sensor; determining the mean of each initial measurement projection and the yaw angle of each sensor detecting the target by using the initial measurement projection of each sensor; determining a straight line q corresponding to each sensor according to the mean of each initial measurement projection and the yaw angle of each sensor detecting the target, and determining a q-z plane corresponding to each sensor according to the straight line q and a z axis; projecting the initialized original measurement data of each sensor onto the q-z plane corresponding to each sensor to obtain preprocessed original measurement data of each sensor; an estimated state information acquisition step: inputting the preprocessed original measurement data of each sensor into a GIW-PHD filtering algorithm to obtain estimated state information of the target detected by each sensor; an estimated state information preprocessing step: preprocessing the estimated state information of the target detected by each sensor to obtain a target state of each sensor; a target state information acquisition step: selecting target states of two sensors with the largest difference in long axis from the target states of each sensor to perform feature extraction and fusion, and obtaining extended state information of a fused target; performing target center position fusion by using the extended state information of the fused target to obtain fused target state information at time k. 2.The feature selection based distributed three-dimensional extended target tracking method according to claim 1, characterized in that, The original measurement data of each sensor is composed of a target measurement source and a Gaussian white noise. 3.The feature selection based distributed three-dimensional extended target tracking method according to claim 1, characterized in that, The yaw angle of each sensor detecting the target is determined by using the initial measurement projection of each sensor, which comprises the following steps: centering the initial measurement projection of each sensor to obtain a centered initial measurement projection of each sensor; calculating a covariance matrix of the centered initial measurement projection of each sensor, and the angle between the eigenvector corresponding to the covariance matrix and the x axis is the yaw angle of each sensor detecting the target. 4.The feature selection based distributed three-dimensional extended target tracking method according to claim 1, characterized in that, The estimated state information acquisition step comprises the following steps: inputting the state and covariance of the target at time k-1 into the GIW-PHD filtering algorithm to obtain a predicted GIW component; inputting the preprocessed original measurement data of each sensor into the GIW-PHD filtering algorithm to update the predicted GIW component and obtain a new GIW component of each sensor; According to the set pruning threshold , a merging threshold between components , and a maximum number of GIW components Prune and merge the updated GIW components of each sensor respectively to obtain new GIW components of each sensor selecting the GIW component with a weight greater than 0.5 from the new GIW component of each sensor to obtain the estimated state information of the target detected by each sensor.

5. The feature selection based distributed three-dimensional extended target tracking method according to claim 3, characterized in that, The estimated state information preprocessing step comprises the following steps: obtaining an elliptical estimated state of an extended target corresponding to each sensor according to the estimated state information of the target detected by each sensor. The yaw angle corresponding to each sensor detecting the target is combined with the elliptical estimated state of the extended target corresponding to each sensor to obtain the target state of each sensor.

Images