Automatic initiation method of radar target track based on B-spline curve fitting

Abstract

The present invention relates to a radar target track automatic initiation method based on B-spline curve fitting, which is applied to a radar data processing system, and includes: receiving radar point track data of the current antenna cycle in real time, matching the point tracks of two adjacent antenna cycles by a pairwise matching method within a wave gate, combining paired point tracks of multiple consecutive cycles into a point track sequence sorted by time, performing motion compensation on the point track sequence according to the motion information of the platform to obtain a compensated absolute motion point track sequence, performing B-spline fitting on all absolute motion point track sequences, storing the fitted curve as a potential track in a potential track library, and finally performing a comprehensive judgment on the starting conditions of the potential tracks, deleting the potential tracks judged to be false, and starting the potential tracks judged to be true as the real tracks. The present invention effectively suppresses false information by utilizing the physical characteristics of the B-spline and the starting condition constraints, thereby improving the starting accuracy and speed.

Description

Technical Field

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

[0002] Track initiation is a primary problem in the field of radar multi-target tracking. At the beginning of track processing, the target is usually far away from the observation position of the radar sensor, resulting in low detection resolution and measurement accuracy. At the same time, in a complex battlefield environment, the occurrence statistics of true and false targets are also irregular. Therefore, an efficient and robust signal processing technique is needed to combine the data accumulated during the extended observation time to achieve a high detection probability and a low false alarm rate. This technique of extending the observation time, in which the target declaration decision is postponed until sufficient information is available, is usually referred to as track-before-detection (TBD) processing. Therefore, the track initiation problem is also treated as a TBD problem.

[0003] At present, the track initiation algorithms commonly used in engineering mainly include sequential processing technology and batch processing technology, represented by the logical method and Hough transform method and its improved algorithm (He You et al., "Radar Data Processing Technology and Application"). Under the background of strong clutter, the improved M / N logical track initiation algorithm can effectively initiate the track, while the virtual track cannot be effectively suppressed. The Hough transform method is combined with multi-scale clustering to solve the problem of target detection and estimation under multiple pseudo-tracks. The improved Hough transform is used to estimate the possible range of the target, and then particle filtering is performed. Although the detection algorithm based on Hough transform can effectively improve the detection probability, it has high complexity and large load, and it is difficult to meet the real-time requirements. Both the logical method and the Hough transform method are performed under the assumption that the target has linear motion. However, in actual situations, due to the presence of a large number of nonlinear maneuvers in the target motion, the difference between the measured point and the predicted position calculated by the algorithm increases. At the same time, it is easy for the actual measurement point to not fall into the tracking gate. Therefore, it is necessary to further study the initiation problem of targets in nonlinear motion.

[0004] The platform radar target initiation faces the following main problems: 1) There are too many false targets that are automatically initiated due to the complexity of the clutter environment and the platform movement; 2) The detection probability of the target itself is not high, or the target motion characteristics are not obvious, resulting in the real target starting speed being too slow; 3) Slow targets with curved motion cannot be started. Summary of the invention

[0005] The present invention aims at the deficiencies of the existing radar automatic start technology and provides a radar target track automatic start method based on B-spline curve fitting, which solves the problem of too many curved motion target starts and false track starts and meets the needs of actual engineering applications.

[0006] The technical solution adopted by the present invention is as follows:

[0007] The automatic initiation method of radar target track based on B-spline curve fitting receives the point track data of multiple antenna cycles; matches the point tracks in the same wave gate in pairs to form a point track sequence in time order; performs platform compensation on the point track sequence to generate an absolute motion point track sequence; performs B-spline fitting on all absolute motion point track sequences to form potential tracks; makes a comprehensive judgment on the starting conditions of the potential tracks; and uses the potential tracks that are judged to be true to start and generate tracks.

[0008] Furthermore, the method for generating an absolute motion point trace sequence includes: receiving the motion data of the platform relative to the earth in real time, the platform motion data including the platform's heading and speed, and recording and storing the platform motion data in real time; for each point trace sequence, taking out the time of each point trace in the time-sorted point trace sequence to obtain a corresponding time series, and using the time series to search for the platform motion information recorded at the corresponding time point; using the motion platform coordinate system at the current moment as the reference coordinate system, converting the coordinates of the historical points at the historical moment to the coordinates of the reference coordinate system at the current moment; that is, according to the platform motion information of the point trace sequence at the corresponding time point obtained by the search, calculating the reference coordinate position of each point trace in the point trace sequence at the current moment using the corresponding time, and obtaining an absolute motion point trace sequence.

[0009] Furthermore, the B-spline fitting method for the absolute motion point trajectory sequence includes: for each time period of the absolute motion point trajectory sequence, selecting a set of spline node vectors, and generating a set of B-spline bases from the node vectors; substituting the data points of the absolute motion point trajectory sequence into the B-spline optimal fitting approximation equation to obtain the optimal approximation fitting curve, and the fitting curve obtained is the potential track corresponding to the absolute motion point trajectory sequence.

[0010] Furthermore, the potential track comprehensive judgment method includes: using two starting conditions to judge the start of the target, the first starting judgment condition: traversing the potential tracks in the potential track library, taking out the point track distance and point track azimuth at each moment of the point track sequence, and taking out the track distance and track azimuth at each moment of the corresponding potential track, and subtracting the point track distance and track distance at the same moment, and subtracting the point track azimuth and track azimuth at the same moment to obtain a distance error sequence and an azimuth error sequence; calculating the root mean square error of the distance error sequence and the azimuth error sequence, and judging the potential track whose root mean square error is outside the radar accuracy range as false, and judging the potential track whose root mean square error is within the radar accuracy range as true; the second starting judgment condition: for the remaining potential tracks judged by the first starting condition, calculating the maximum curvature of the potential track, and inferring the corresponding overload from the calculated maximum curvature, judging the potential track whose overload exceeds the threshold as false, and judging the potential track whose overload does not exceed the threshold as true.

[0011] Beneficial effects of the present invention:

[0012] (1) The traditional logic method and Hough transform method cannot start for curved moving targets, while the radar target track automatic start method based on B-spline curve fitting can start not only for linear moving targets, but also for curved moving targets;

[0013] (2) Compared with other methods, the radar target track automatic initiation method based on B-spline curve fitting innovatively utilizes the radar detection accuracy and the physical characteristics of target motion dynamics to eliminate false tracks, thereby increasing the false track suppression rate by 50%. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 The figure is a flow chart of the automatic initiation method of radar target track based on B-spline curve fitting.

[0015] Figure 2 Schematic diagram of B-spline curve fitting track. DETAILED DESCRIPTION

[0016] The radar target track automatic initiation method based on B-spline curve fitting proposed by the present invention firstly uses the real-time heading and speed of the moving platform to make motion compensation for the point track data at each moment, eliminating the influence of the platform movement on the track initiation; secondly, adopts the B-spline curve interpolation algorithm to realize nonlinear dynamic programming, extracts the B-spline curve features to find the optimal target, and expands the scope of the traditional track automatic initiation from the linear motion target to the curve motion, and widely adapts to the automatic initiation of targets of different types and different motion characteristics in real scenes; finally, the radar detection accuracy and the physical characteristics of the target motion dynamics are used to realize pruning in the track generation, and the problem of track branch scale explosion is effectively controlled in actual engineering applications. The pruning method extracts the legal targets in the entire track space while eliminating false targets. Due to the locality of the B-spline curve, a more universal curve fitting is realized; the smoothness of the B-spline curve determines that the speed and acceleration of the track are continuous, strictly limiting the generation of false targets. Thereby ensuring the characteristics of high precision, low false rate and fast speed of track initiation.

[0017] The specific implementation steps of the present invention are as follows Figure 1 The detailed description is as follows:

[0018] 1. Initialization. Create a memory to store the trace sequence and initialize the storage unit. Store the real-time received radar trace data in real time. k The point trace received at the moment is recorded as p(t k ).

[0019] 2. Perform gate-in-wave matching on the traces of the current antenna cycle and the traces of the historical adjacent cycles, and combine the paired traces of multiple consecutive antenna cycles into a trace sequence sorted by time, recorded as p(t1:t k ). The gate G size is calculated according to the following formula:

[0020] G=[v min ,v max ]×t (1)

[0021] Among them, v min is the minimum velocity required to start, v max is the maximum starting speed required, and t is the time interval between two adjacent antenna cycles.

[0022] 3. Perform motion compensation on the point trace sequence according to the platform's motion information to obtain the compensated absolute motion point trace sequence: receive the platform's motion data relative to the ground in real time (the platform's heading φ at time t ship (t) and speed v ship (t)), taking the motion platform as the reference coordinate system, the heading and speed of the platform are converted into the reference rectangular coordinate system. The speed of motion in the x and y directions is:

[0023]

[0024] The platform motion data is recorded and stored in real time. Take the point trace sequence p(t1:t k ), the time series of each point is t1:t k , then the platform motion information corresponding to these times can be approximated as (v x (t1:t k ), v y (t1:t k )), and use the motion information to perform motion compensation on the point sequence to obtain the absolute motion point sequence Z(t1:t k ), that is, for t1:t k Each point trace at a time point is compensated as follows:

[0025]

[0026] 4. For all absolute motion point sequences Z(t1:t k ) performs B-spline fitting, and the fitted curve is stored in the potential track library as a potential track. Fitting defines a cubic B-spline curve:

[0027]

[0028] Among them, B i,3 (t) is called a cubic B-spline piecewise function. A spline curve consists of a series of piecewise functions.

[0029] For the spline basis function, the following recursive definition form is used for calculation:

[0030]

[0031] B i,γ The i in t represents the sequence number, and γ represents the number of times, and it is agreed that 0 / 0 = 0. i ,t i+1 …t i+γ A total of γ+1 nodes determine B i,γ . The interval [t i ,t i+γ ] is B i,γ support range.

[0032] For a set of absolute motion points Z(t1:t w ),t1:t w ∈[a,b],t1:t w Represents a set of node vectors given on the interval [a,b].

[0033] The specific steps of track fitting are as follows:

[0034] Step 4.1: Select a set of spline knot vectors And ensure that a=t0 <t0<t1<…<t N = b;

[0035] Step 4.2: Substitute this set of node vectors into the recursive relation (5) to obtain B i,3 (t);

[0036] Step 4.3: Corresponding data point B i,3 Substitute (t) into the matrix M:

[0037]

[0038] Step 4.4: Solve the best fit approximation equation M for discrete data points T Mβ=M T Z.

[0039] The specific form of this equation is:

[0040]

[0041] The solution to the equation is β=M T Z;

[0042] Step 4.5: Calculation Get the spline curve;

[0043] Step 4.6: Take t1:t according to the calculated spline curve as the fitting track. w The track coordinates at the time point are stored as potential tracks in the potential track library X(t1:t w ).

[0044] 5. Finally, the starting conditions of the potential tracks are comprehensively judged. Potential tracks judged as false are deleted, and potential tracks judged as true are taken as the starting points of the real tracks.

[0045] The legitimacy of the track should usually meet certain physical constraints. This method mainly considers two steps of judgment conditions, one is the measurement error condition judgment, and the second is the dynamic overload condition judgment.

[0046] Step 5.1 Radar measurement error constraint determination

[0047] The radar measurement error is defined as the standard deviation of the distance dis , the standard deviation of the azimuth error is σ azi . Define the measurement standard deviation vector △=[3σ dis ,3σ azi The conversion process of radar system measurement coordinate system into Cartesian rectangular coordinate system is as follows:

[0048]

[0049] Where ρ is the distance, is the azimuth. Then the measurement error of the Cartesian coordinate system can be expressed by the Jacobian matrix J of the above transformation matrix T. T As the error transformation matrix, that is,

[0050] δ th =J T △ (8)

[0051] The error δ obtained th , as a constraint condition for judgment. Each moment t1:t w The point distance Dis dot (t1:t w ) and point position Azi dot (t1:t w ) is taken out. Calculate the corresponding potential track X(t1:t w ) at each moment of the track distance Dis trk (t1:t w ) and track bearing Azi trk (t1:t w ). Calculate the error between the point track distance and the track distance at the same time, and the error between the point track azimuth and the track azimuth at the same time, and get the distance error sequence e dis (t1:tw ) and the azimuth error sequence e azi (t1:t w ).

[0052] The root mean square error [RMSE] of the distance error sequence and the azimuth error sequence is calculated. dis ,RMSE azi ], the root mean square error and δ th By comparison, potential tracks whose RMS errors are outside the radar accuracy range are judged as false, and potential tracks whose RMS errors are within the radar accuracy range are judged as true.

[0053] Step 5.2 Determination of overload constraint conditions

[0054] The parametric equations of each segment of the cubic B-spline curve obtained in step 4 are expressed as:

[0055]

[0056] According to the parametric equation of the cubic B-spline curve, its derivative is calculated as:

[0057]

[0058]

[0059] Calculate the curvature of the curve in the plane

[0060]

[0061] Calculate the track motion overload as

[0062]

[0063] Where V is the overload in the direction of the target's motion speed, and n trajectory Is it greater than the set threshold? If it is greater than the threshold, the potential track is judged to be false; if it is less than the threshold, the potential track is judged to be true.

Claims

1. A radar target track automatic initiation method based on B-spline curve fitting, characterized in that: Receive the point trace data of multiple antenna cycles; match the point traces in the same wave gate in pairs to form a point trace sequence in time order; perform platform compensation on the point trace sequence to generate an absolute motion point trace sequence; Perform B-spline fitting on all absolute motion point trace sequences to form potential tracks, including: for each absolute motion point trace sequence Z(t1:t w ),t1:t w ∈[a,b] time period t1:t w , t1:t w Represents a set of knot vectors given on the interval [a,b]; select a set of spline knot vectors And ensure that a=t0<t0<t1<…<t N = b, a set of B-spline bases B is generated by the node vectors i,3 (t), the data points of the absolute motion trace sequence, that is, the B-spline basis B i,3 (t), substitute into the B-spline optimal fitting approximation equation M T Mβ=M T Z, where The solution to the equation for the best approximation fitting curve is β = M T Z, the fitting curve obtained That is, the potential track corresponding to the absolute motion point track sequence; making a comprehensive judgment on the starting conditions of the potential track; and starting and generating the track using the potential track that is judged to be true.

2. The radar target track automatic initiation method based on B-spline curve fitting according to claim 1, characterized in that: The method for generating an absolute motion point trace sequence comprises: receiving in real time the motion data of a platform relative to the earth, the platform motion data including the heading and speed of the platform, and recording and storing the platform motion data in real time; for each point trace sequence, taking out the time of each point trace in the point trace sequence sorted by time, obtaining a corresponding time sequence, and using the time sequence to search for platform motion information recorded at a corresponding time point; using the motion platform coordinate system at the current moment as a reference coordinate system, converting the coordinates of historical points at historical moments into the coordinates of the reference coordinate system at the current moment; that is, according to the platform motion information of the point trace sequence at the corresponding time point obtained by searching, calculating the reference coordinate position of each point trace in the point trace sequence at the current moment using the corresponding time, and obtaining an absolute motion point trace sequence.

3. The radar target track automatic initiation method based on B-spline curve fitting according to claim 1, characterized in that: The potential track comprehensive judgment method comprises: using two starting conditions to judge the start of the target, the first starting judgment condition: traversing the potential tracks in the potential track library, taking out the point track distance and the point track azimuth at each moment of the point track sequence, and taking out the track distance and the track azimuth at each moment of the corresponding potential track, and subtracting the point track distance and the track distance at the same moment, and subtracting the point track azimuth and the track azimuth at the same moment, to obtain the distance error sequence and the azimuth error sequence; The root mean square error of the distance error sequence and the azimuth error sequence is calculated. The potential track whose root mean square error is outside the radar accuracy range is judged as false, and the potential track whose root mean square error is within the radar accuracy range is judged as true. The second starting judgment condition: for the remaining potential tracks judged by the first starting condition, the maximum curvature of the potential track is calculated, and the corresponding overload is inferred from the calculated maximum curvature. The potential track whose overload exceeds the threshold is judged as false, and the potential track whose overload does not exceed the threshold is judged as true.

Images