Automatic initiation method for radar target based on cross-period target morphology quality consistency inspection

Abstract

The present invention discloses a radar target automatic initiation method for cross-cycle target morphology quality consistency inspection, which forms a point trace data pool with current cycle data and real-time radar point trace data within a time window period and stores them by period. In the window point trace pool, two point traces of different periods are sampled according to the criterion of maximum period interval to initialize the target model, and the predicted point position of the target model at each time stamp in the time window is obtained, and a corresponding measurement wave gate is established with the position, and then other supporting point traces of the target model are searched in the measurement wave gate of each time stamp point in the window point trace pool. Finally, all target support points are fitted by the least squares method, and the consistency test function is used to determine whether the target model is a real moving target. The above sampling and judgment process is repeated until the point traces with the maximum period interval are sampled. This method can minimize the detection root mean square error, and at the same time, the proprietary wave gate of each different time stamp is calculated to reduce the amount of calculation.

Description

Technical Field

[0001] The present invention relates to radar data processing technology in the field of radar technology, and in particular to a shipborne radar target automatic initiation technology. Background Art

[0002] The large number of random clutter points in complex environments has a serious impact on target detection, posing a significant challenge to the detection capabilities of modern radars. Track-before-detection (TBD) technology is an effective method for detecting small, moving, and weak targets with low signal-to-noise ratios. Its essence is to leverage the continuity of the target's motion state in space, estimate the possible target trajectory, and perform non-coherent accumulation of multiple frames of data to improve the signal-to-noise ratio (SNR) of detection, thereby detecting the true target. Since its introduction, TBD technology has achieved significant development and widespread application in the field of weak target detection. Researchers have proposed various implementation methods for TBD technology, including those based on the Hough transform, multi-level hypothesis testing, particle filtering, and dynamic programming. The original Hough transform technology is based on a Cartesian coordinate system. When applied to target detection in search radars, the radar measurement coordinates need to be converted into Cartesian coordinates for processing. Subsequently, a polar coordinate Hough transform method was proposed to facilitate the processing of radar data. This method directly uses the radial range and azimuth of the radar data for Hough transform, without converting the echo data into a Cartesian coordinate system. This method effectively detects and tracks targets with linear trajectories within the radar's detection area, and exhibits robust detection and tracking performance for targets moving linearly in three or higher-dimensional space. The TBD algorithm, based on multi-level hypothesis testing, organizes all possible target trajectories in a tree format. The tree is updated, managed, and pruned with each frame of data, while a sequential probability ratio test is used for judgment. This method is relatively simple in principle, highly convenient, and suitable for engineering applications. However, in low signal-to-noise ratio (SNR) conditions, a large number of candidate trajectory starting points is required to reduce the probability of missed detection. This rapidly increases the number of branches in the tree, resulting in a combinatorial explosion, rapidly increasing the algorithm's computational complexity, and complicating the correlations, severely impacting algorithm performance. In the particle filter-based track-before-detect (PF-TBD) algorithm, a discrete vector is used to describe the presence or absence of a target, and its state transition probabilities are calculated using a Markov stationary random process. It is then used together with the assumed target state vector to form a composite state vector, and their mixed estimation is achieved through the PF algorithm. The researchers then used different target models and noise models to make different improvements. In order to improve the probability of target detection under low signal-to-noise ratio, the TBD algorithm based on dynamic programming uses multi-stage decision optimization to realize trajectory search for weak targets. This requires the selection of the value of a certain variable at each stage. The selected value must meet the requirements of enabling the entire process to achieve the optimal result according to the given criteria, so that the global optimality can be approximated by a single-step optimal approach. Due to the advantages of the TBD algorithm based on dynamic programming, which has clear principles and easy hardware implementation, it is widely used in scenarios with certain prior information and has also achieved many results in engineering.However, under low signal-to-noise ratio, the target detection performance of the dynamic programming algorithm cannot be improved no matter how the frame number is increased.

[0003] Radar target initiation in complex environments faces the following main problems: 1) Too many clutter points affect the judgment of normal target hypothesis testing, resulting in too many false tracks being automatically initiated; 2) Too much clutter causes a rapid increase in computational complexity, seriously degrading algorithm performance. Summary of the Invention

[0004] In view of the shortcomings of existing radar automatic initiation technology, the present invention provides a radar target automatic initiation method for cross-cycle target morphology quality consistency inspection, which solves the problems of too many false track starts and excessive calculation, and meets the needs of actual engineering applications.

[0005] The present invention proposes a radar target automatic initiation method for cross-period target morphology quality consistency inspection. It assumes that the target motion trajectory is linear within a certain time and space range. Through the real-time acquisition of radar trace sequence, the traces within the time window are stored. The characteristic of linear trajectory that only two points of data are needed to determine the parameters is utilized. A two-point maximum time interval sampling consistency algorithm is designed within the framework of hypothesis testing. In the hypothesis sampling stage, two traces are sampled as trace pair data for hypothesis parameter estimation. In a heavy clutter environment, the combined sampling method has more advantages than the random sampling method. The optimal estimation time of the trace sequence is obtained under the criterion of minimum root mean square error based on consistency estimation. In the inspection stage under this sampling method, according to the different periods of the data in the time series, due to the residual distribution characteristics of the two-point estimation, the optimal wave gate of different period inspection is adopted to ensure that the target can fall into the wave gate while eliminating clutter as much as possible, ensuring the detection probability of the target, and the points falling into the wave gate are regarded as target inliers. Finally, all target inliers are fitted by the least squares method, and the points with consistency inspection function greater than the threshold are output as true targets. The specific technical solution includes:

[0006] The radar trace data is acquired in real time, and the current period data and the radar trace data within the time window period are combined into a trace data pool, and the data in the trace pool are stored by period. The linear Gaussian process is used to model the target motion, and two traces of different periods are sampled in the window trace pool according to the maximum period interval criterion to initialize the target model. The predicted point position of the target model at each time stamp in the time window is obtained, and the corresponding time-related measurement gate is established with this position. Then, other supporting traces of the target model are searched in the measurement gate of each time stamp point in the window trace pool. Finally, all target support points are fitted by the least squares method, and the target morphological quality is calculated using the consistency test function. The morphological quality is used to determine whether the established target model is a real moving target. The above sampling and judgment process is repeated until the traces with the maximum period interval are sampled.

[0007] Furthermore, the time-correlated measurement gate includes: assuming that the variance σ of the radar measurement error is within the range of the time window size w, each period k, k∈{k1,k2,k3,...,k w The measurement gate of} is in G is the gate coefficient.

[0008] Furthermore, the consistency check function includes: passing the traces within the time window size w through the wave gate to obtain all the traces within the wave gate, assuming that the time window k1 to k w The sum of all traces within the wave gate within the range is N, and the consistency test function of the corresponding target model is in:

[0009]

[0010]

[0011] where m i (k) is the fitting position of the target model in the kth cycle, x i (k) is the point position closest to the fitting value in the wave gate of the kth cycle, z i (k) is the point position in the wave gate of the kth cycle; the consistency test function of the target model is used as the morphological quality. If the morphological quality is greater than the threshold, it is determined to be a real target, otherwise it is considered false.

[0012] The present invention can utilize the situational characteristics of the target and the estimation of radar measurement errors to finely process the clutter traces inside and outside the wave gate, effectively suppress false detection, thereby overcoming the influence of clutter on moving target detection, improving the moving target detection performance in heavy clutter areas, and greatly reducing the computational cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 This is a preferred flow chart of the present invention.

[0014] Figure 2 Schematic diagram of wave gates with different periods within the window. DETAILED DESCRIPTION

[0015] The present invention will be further explained below with reference to the accompanying drawings and preferred embodiments.

[0016] The preferred implementation steps of the present invention are as follows Figure 1 As shown, the description is as follows:

[0017] Step 1: Get radar trace data in real time, combine the current period data and the radar trace data in the time window period into a trace data pool, and the data in the trace pool is divided into periods k1:kw Storage, represented as o i Represents the set of all traces in the corresponding period.

[0018] Step 2: Sample a point in the k1th period as p(k1), and sample the kth w A point in the cycle is p(k w ), use these two points to initialize a linear motion model as a moving target hypothesis, that is, the position of the moving target in the kth period in the time window is:

[0019]

[0020] in

[0021] Step 3: Calculate the measurement detection gate of the moving target for each cycle, assuming that the variance of the radar measurement error σ is w In the range of , the measurement gate of each cycle k is Among them, G is the wave gate coefficient, such as Figure 2 shown.

[0022] Step 4: In the trace pool, set the time window k1:k w Point P inside w After the wave gate screening in the previous step, the traces inside the wave gate are obtained, that is, the traces inside the wave gate of the kth cycle z(k), the time window k1:k w The total number of points within all wave gates is N.

[0023] Step 5: Search for other supporting traces of the target model in the measurement wave gate of each cycle in the window trace pool, that is, select the target position corresponding to the kth cycle in the trace o(k) corresponding to each cycle k. The nearest point is the support point of the target, denoted as x(k), with a time window of k1:k w The total number of supporting points in is n.

[0024]

[0025] Step 6: Set the time window k1:k w All supporting points in the range of x(k1:k w ), a fitted straight line trajectory is obtained by least squares straight line fitting, and the fitting position m(k) of the trajectory corresponding to the kth period is obtained.

[0026] Step 7: Make a consistency judgment on the target model. The consistency check function of the target model is in,

[0027]

[0028]

[0029] Step 8: The consistency check function of the corresponding target model is used as the morphological quality. If the morphological quality is greater than the threshold, it is determined to be a real target; otherwise, it is considered a false target. The threshold is set as Q / w, where Q is the evaluation coefficient, usually greater than 50, and w is the time window size.

[0030] Step 9: Repeat the sampling and judgment process from step 2 to step 8 until the trace of the k1th cycle and the kth cycle are w All traces of the period are completely sampled, ending the operation of the current period.

Claims

1. A radar target automatic initiation method for cross-period target morphology quality consistency inspection, characterized by: The radar trace data is acquired in real time, and the current period data and the radar trace data within the time window period are combined into a trace data pool. The data in the trace pool is stored periodically. The target motion is modeled using a linear Gaussian process. Two traces of different periods are sampled in the window trace pool according to the maximum period interval criterion to initialize the target model. The predicted point position of the target model at each time stamp in the time window is obtained, and a time-related measurement gate corresponding to the position is established. Then, other supporting traces of the target model are searched in the measurement gate of each time stamp point in the window trace pool. Finally, all target support points are fitted by the least squares method, and the target morphological quality is calculated using the consistency test function. The morphological quality is used to determine whether the established target model is a real moving target. The above sampling and judgment process is repeated until the traces with the maximum period interval are sampled. The consistency check function includes: passing the traces within the time window size w through the wave gate to obtain all the traces within the wave gate, assuming that the time window k1 to k w The sum of all traces within the wave gate within the range is N, and the consistency test function of the corresponding target model is in: where m i (k) is the fitting position of the target model in the kth cycle, x i (k) is the point position closest to the fitting value in the wave gate of the kth cycle, z i (k) is the point position in the wave gate of the kth cycle; the consistency test function of the target model is used as the morphological quality. If the morphological quality is greater than the threshold, it is determined to be a real target, otherwise it is considered false.

2. The radar target automatic initiation method for cross-period target morphology quality consistency inspection according to claim 1 is characterized in that: The time-correlated measurement gate includes: assuming that the variance of the radar measurement error σ, within the range of the time window size w, each period k, k∈{k1,k2,k3,...,k w The measurement gate of} is in G is the gate coefficient.

Images