Improved track initiation method based on correction Hough transformation

Abstract

The invention discloses an improved track initiation method based on correction Hough transformation, and designs a method for filtering clutters first and then initiating the track in order to solve the problem that the false alarm rate of the track initiation is high in a strong clutter environment. The calculation amount of the algorithm is reduced by using a clutter filtering algorithm, clutters are few and relatively concentrated after clutter filtering, and a fine track initiation algorithm suitable for an environment with few clutters and relatively concentrated is designed, so that the calculation amount and storage space of a system occupied by track initiation are reduced as much as possible, and the track initiation efficiency is improved. And high-quality track initiation is ensured. The track initiation algorithm is provided.

Description

Technical Field

[0001] The invention relates to the technical field of radar data processing, and in particular to an improved track initiation method based on modified Hough transform. Background Art

[0002] The purpose of track initiation is to quickly and accurately establish a track profile of the real target, paving the way for subsequent tracking and maintenance. Therefore, track initiation is the primary issue in track processing. For multi-target track processing, track initiation is the first step and the foundation for track tracking. Since detected targets generally appear from far to near within the effective detection range of the search radar, at the initial track initiation, the target is far from the radar, limited by the radar's low resolution and poor measurement accuracy. Furthermore, there is no true statistical pattern in the appearance of true and false targets. Therefore, track initiation is a difficult problem in multi-target track processing.

[0003] Existing track initiation algorithms can be divided into two categories: sequential processing technology and batch processing technology. Generally, sequential data processing technology is suitable for track initiation in weak clutter environments, mainly including heuristic rule methods, logic-based methods, and modified logic methods; batch processing technology is more suitable for strong clutter environments, mainly including standard Hough transform, modified Hough transform, sequence Hough transform and other methods. The logic-based method has a better effect in initiating tracks when the probability of false alarm is relatively low, but there are more false tracks in complex environments; the Hough transform rule is suitable for environments where the tracks are straight in a strong clutter background, but the Hough transform method usually requires multiple scans to better initiate the track, and the large amount of calculation does not meet the needs of engineering applications. This patent conducts in-depth research on this issue and proposes an improved track initiation algorithm. Compared with the traditional track initiation algorithm, the algorithm in this paper has better starting effect and running speed, and has strong engineering application value. Summary of the Invention

[0004] The object of the present invention is to provide an improved track initiation method based on modified Hough transform to solve the problems raised in the above background technology.

[0005] In order to achieve the above object, the present invention adopts the following technical solution: an improved track initiation method based on modified Hough transform, comprising:

[0006] Step S100: Extract the first measurement point d from the measurement set D i =(x i ,y i , Tab i ), then extract a Tab that is larger than the first measurement point Tab iThe measurement points are determined, the coordinates of the two points are determined, and the period of the two measurement points is calculated.

[0007] Step S200: Determine whether the distance threshold and speed threshold between two measurement points are met;

[0008] Step S300: Calculate the parameters of the straight line between the two points and store the data results in the track parameter set P;

[0009] Step S400: traverse the track parameter set P, accumulate and vote for p = (ρ, θ) in the parameter space, select the peak point, and match all points that satisfy d ≤ ε l The measurement points in the track parameter set P are filtered through time series to remove those with discontinuous or isolated time series labels, retaining continuous points. Then, a threshold is used to select track parameters with more than m measurement points in the track parameter set P. Step S500: The measurement data in the track parameter P are traversed to calculate their positional relationships and determine whether they meet the operating rules. If they meet the rules, they are added to the real track set. Otherwise, they are considered false tracks and added to the false track set.

[0010] Step S600: The actual track parameter set stores the starting track, and the track initiation is completed.

[0011] Furthermore, the specific process of step S100 includes:

[0012] Step S101: Extract the first measurement point d from the measurement set D i =(x i ,y i , Tab i ), where the set of measurements to be measured is D = [D t , Tab t ], including all the measured points, D t Indicates the coordinates of the measurement point (x t ,y t ), Tab t Indicates the measurement cycle number of the measurement point, cycle number = (1, 2, ..., N);

[0013] Step S102: Extract a Tab that is larger than the first measurement point Tab i The measurement point is associated with the measurement point extracted in step S101. Assume that the selected point is d j , then the two points are d i =(x i ,y i ) and d j =(x j ,y j ), the period between the two measurement points is Δt=(Tab j -Tabi )×T, where T is the radar scanning period.

[0014] Furthermore, the specific process of step 200 includes:

[0015] Step 201: Calculate the distance between two measurement points D is obtained by calculation or estimation init , if R≤D init , indicating that the distance between the two points is within the set distance threshold D init Inside;

[0016] Step 202: Further determine the velocity between the two points and obtain V by calculation or estimation. min 、V max , according to the principle of kinematics, calculate the speed V=R / Δt, if V min <V <V max , which means that the speed threshold is satisfied between the two measurement points, and the process continues to step S300. Otherwise, the process goes to step S102 and a new point d is extracted. j .

[0017] Furthermore, the specific process of step S300 includes:

[0018] Calculate the straight line parameter p between two points according to the two-point straight line formula i =(ρ i ,θ i ), the data results are stored in the track parameter set P, where the track parameter set P = [p t , r t ] is a dynamic linked list, where p t Used to store the linear parameters (ρ t ,θ t );x t The points in the xy plane that satisfy the straight line parameters and the label parameters of the measurement cycle to which they belong (x t ,y t , Tab t ); Initialize the track parameter set P = NULL, set the threshold to τ, and the minimum number of starting points of the track to m.

[0019] Furthermore, the specific process of step S400 includes:

[0020] Step S401: Traverse the track parameter set P, accumulate and vote for p = (ρ, θ) in the parameter space, and find small areas where more points fall. The center points of these small areas are the peak points.

[0021] Step S402: The peak point corresponds to a straight line in the xy plane, matching all lines that satisfy d≤ε lThe measurement point, where d is the shortest distance from the measurement point to the straight line, ε l The preset threshold for permissible;

[0022] Step S403: By using time series filtering, the measurement points with discontinuous and isolated time series labels in the track parameter set P are removed, and the continuous points are retained. t Update to the retained consecutive points;

[0023] Step S404: retaining the track parameters with the number of measurement points greater than m in the track parameter set P through threshold selection.

[0024] Furthermore, the specific process of step S500 includes:

[0025] Step S501: Based on the measurement cycle number Tab, the measurement data in the track parameter P can be traversed to calculate their positional relationships and determine whether they meet the operating rules. Based on the time series parameter Tab in the track data set P, each track point is sorted in descending order (1, 2...N) to construct a time series coordinate matrix: d k+1 For one of the measurement points, use the measurement points d of the first two measurement cycles k-l (x k-1 ,y k-1 ) and d k (x k ,y k ), calculate the velocity components:

[0026]

[0027] Step S502: Predict the position of the k+1th cycle based on the current speed:

[0028]

[0029] The predicted position and the actual measurement point d k+1 Compare the position residuals and calculate the residuals:

[0030]

[0031] Step S503: Determine whether it satisfies the dynamic consistency and velocity smoothness. The dynamic consistency requires the residual ε≤ε th , where ε th is the maximum allowed prediction error; velocity smoothness requires that the velocity change rate (i.e., acceleration) of adjacent cycles does not exceed the threshold:

[0032] where a max is the target maximum acceleration.

[0033] Step S504: If all periodic measurement points meet the above conditions, the track is retained, duplicate points are removed, and it is added to the real track set. If any period does not meet the motion law, the track is moved to the false track set.

[0034] Compared with the prior art, the present invention has the following advantages:

[0035] (1) In a strong clutter environment, although the standard Hough transform-based initiation method successfully initiated four tracks, a large number of false tracks appeared, and there was a serious track clustering phenomenon; the modified Hough transform only initiated three tracks; while the initiation method based on the improved modified Hough transform successfully initiated four tracks, effectively overcoming the influence of strong clutter on the data, and having a good initiation effect.

[0036] (2) As the clutter level increases, the time required for the track initiation algorithm increases. The improved modified Hough transform method is inferior to the traditional algorithm in terms of real-time performance. However, the false alarm probability is much lower than that of the traditional algorithm. In a strong clutter environment, the algorithm proposed in this patent is better. This shows that although this method sacrifices some real-time performance, it achieves a lower false alarm rate and better track initiation effect.

[0037] The present invention will be further described below with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 Flowchart of the present invention.

[0039] Figure 2 It is a situation diagram of clutter points and real points.

[0040] Figure 3 This is the experimental result of the traditional Hough transform method.

[0041] Figure 4 This is the experimental result of the modified Hough transform method.

[0042] Figure 5 This is the experimental result of the improved Hough transform method. DETAILED DESCRIPTION

[0043] Combine Figure 1 , an improved track initiation method based on modified Hough transform: In this method, there are a set of measurements to be measured and a track parameter set.

[0044] The measurement set to be measured D=[D t , Tab t ], including all the measured points, D t Indicates the coordinates of the measurement point (x t,y t ), Tab t Indicates the measurement cycle number of the measurement point, cycle number = (1, 2, ..., N).

[0045] Track parameter set P = [p t , r t ] is a dynamic linked list, where p t Used to store the linear parameters (ρ t ,θ t );r t The points in the xy plane that satisfy the straight line parameters and the label parameters of the measurement cycle to which they belong (x t ,y t , Tab t ); Initialize the track parameter set P = NULL, set the threshold to τ, and the minimum number of starting points of the track to m;

[0046] Step S100: Extract the first measurement point d from the measurement set D i =(x i ,y i , Tab i ), then extract a Tab that is larger than the first measurement point Tab i The measurement points are determined, the coordinates of the two points are determined, and the period of the two measurement points is calculated.

[0047] Step S100 includes the following steps:

[0048] Step S101: Extract the first measurement point d from the measurement set D i =(x i ,y i , Tab i ), where the set of measurements to be measured is D = [D t , Tab t ], including all the measured points, D t Indicates the coordinates of the measurement point (x t ,y t ), Tab t Indicates the measurement cycle number of the measurement point, cycle number = (1, 2, ..., N);

[0049] Step S102: Extract a Tab that is larger than the first measurement point Tab i The measurement point is associated with the measurement point extracted in step S101. Assume that the selected point is d j , then the two points are d i =(x i ,y i ) and d j =(x j,y j ), the period between the two measurement points is Δt=(Tab j -Tab i )×T, where T is the radar scanning period.

[0050] Step S200: Determine whether the distance threshold and speed threshold between two measurement points are met;

[0051] Step S200 includes the following steps:

[0052] Step 201: Calculate the distance between two measurement points D is obtained by calculation or estimation init , if R≤D init , indicating that the distance between the two points is within the set distance threshold D init Inside;

[0053] Step 202: Further determine the velocity between the two points and obtain V by calculation or estimation. min 、V max , according to the principle of kinematics, calculate the speed V=R / Δt, if V min <V <V max , which means that the speed threshold is satisfied between the two measurement points, and the process continues to step S300. Otherwise, the process goes to step S102 and a new point d is extracted. j .

[0054] Step S300: Calculate the straight line parameter p between two points according to the straight line two-point formula i =(ρ i ,θ i ), and store the data results in the track parameter set P.

[0055] Step S400: traverse the track parameter set P, accumulate and vote for p = (ρ, θ) in the parameter space, select the peak point, and match all points that satisfy d ≤ ε l The measurement points with discontinuous and isolated time series labels in the track parameter set P are removed through time series filtering, and the continuous points are retained. Then, the track parameters with more than m measurement points in the track parameter set P are retained through threshold selection.

[0056] Step S400 includes:

[0057] Step S401: In step S300, many (ρ, θ) are obtained. In the parameter space, they will intersect. The more intersections there are, the more likely they are the tracks of the same target in the xy plane. Traverse the track parameter set P, accumulate and vote for p = (ρ, θ) in the parameter space. Each line parameter corresponds to an isolated point in the parameter space. Discretely divide the parameter space p = (ρ, θ) into several small areas, and find the small areas that fall into more points. The center points of these small areas are the peak points.

[0058] Step S402: The peak point corresponds to a straight line in the xy plane, matching all lines that satisfy d≤ε l The measurement point, where d is the shortest distance from the measurement point to the straight line, ε l The preset threshold for permissible;

[0059] Step S403: By filtering the time series, remove the discontinuous and isolated measurement points in the track parameter set P, retain the continuous points, and t Update to the retained consecutive points;

[0060] Step S404: retaining the track parameters with the number of measurement points greater than m in the track parameter set P through threshold selection.

[0061] Step S500: traverse the measurement data in the track parameter P to calculate its position relationship and determine whether it meets the operation rules. If it meets the rules, it is added to the real track set; otherwise, it is considered a false track and added to the false track set.

[0062] Step S500 includes:

[0063] Step S501: Based on the measurement cycle number Tab, the measurement data in the track parameter P can be traversed to calculate their positional relationships and determine whether they meet the operating rules. Based on the time series parameter Tab in the track data set P, each track point is sorted in descending order (1, 2...N) to construct a time series coordinate matrix: d k+1 For one of the measurement points, use the measurement points d of the first two measurement cycles k-1 (x k-1 ,y k-1 ) and d k (x k ,y k ), calculate the velocity components:

[0064]

[0065] Step S502: Predict the position of the k+1th cycle based on the current speed:

[0066]

[0067] The predicted position and the actual measurement point d k+1 Compare the position residuals and calculate the residuals:

[0068]

[0069] Step S503: Determine whether it satisfies the dynamic consistency and velocity smoothness. The dynamic consistency requires the residual ε≤ε th , where ε th is the maximum allowed prediction error; velocity smoothness requires that the velocity change rate of adjacent cycles, that is, the acceleration, does not exceed the threshold:

[0070] where a max is the target maximum acceleration.

[0071] Step S504: If all periodic measurement points meet the above conditions, the track is retained, duplicate points are removed, and it is added to the real track set. If any period does not meet the motion law, the track is moved to the false track set.

[0072] Step S600: The actual track parameter set stores the starting track, and the track initiation is completed.

[0073] Example: This patent mainly compares the traditional Hough transform algorithm and evaluates the algorithm performance based on the success probability of the initial target track, the probability of the initial false track, and the initial running time of multiple target tracks.

[0074] In order to more intuitively compare the performance of various track initiation algorithms, the standard Hough transform method, the modified Hough transform method, and the improved modified Hough transform method are experimentally simulated in the same simulation environment. Taking multi-target track initiation as an example, it is assumed that there are four targets in the monitoring area moving in a uniform straight line, and the initial values ​​are:

[0075] Target 1: x1 = 40 km, y1 = 20 km, vx1 = 0.3 km / s, vy1 = -0.18 km / s;

[0076] Target 2: x2 = 20 km, yl = 80 km, vxl = 0.3 km / s, vy1 = -0.18 km / s;

[0077] Target 3: x2 = 60 km, yl = 40 km, vx1 = 0.3 km / s, vy1 = -0.18 km / s;

[0078] Target 4: x2 = 80 km, y1 = 70 km, vx1 = 0.3 km / s, vyl = -0.18 km / s;

[0079] Sampling period Ts = 4s; the measurement noise has a mean of 0 and a variance of σ x =σ y = 0.05 km of Gaussian white noise; allowable error Δθ = π / 100, Δρ = 6σx; the number of clutter per cycle follows a Poisson distribution, and by changing the parameters, clutter of varying intensities can be obtained. The speed thresholds set in the improved Hough transform track initiation method proposed in this patent are: maximum speed Vmax = 0.7 km / s, minimum speed Vmin = 0.05 km / s.

[0080] Taking λ=100, the situation diagram of the clutter points and the real target points generated in three consecutive scanning cycles is shown in the figure. In the figure, "*" represents the point of the first scan, "+" represents the point of the second scan, "Δ" represents the point of the third scan, and "○" represents the real target track.

[0081] The comparative experimental results in this paper are as follows Figures 2 to 4 Shown: Comparison Figures 3-5 It can be seen that in a strong clutter environment, although the standard Hough transform-based track initiation method successfully initiated four tracks, a large number of false tracks appeared, and there was a serious track clustering phenomenon; the modified Hough transform only initiated three tracks; and the initiation method based on the improved modified Hough transform successfully initiated four tracks, effectively overcoming the influence of strong clutter on the data, and having a good initiation effect.

[0082] As can be seen from Tables 1 and 2 below, the track initiation algorithm takes longer as the clutter level increases. The improved modified Hough transform method is inferior to the traditional algorithm in terms of real-time performance. However, the false alarm rate is much lower than that of the traditional algorithm. In strong clutter environments, the algorithm proposed in this patent performs better. This shows that although this algorithm sacrifices some real-time performance, it achieves a lower false alarm rate and better track initiation results.

[0083] Table 1 Statistics of average running time (s) of traditional Hough transform method and improved modified Hough transform method

[0084]

[0085] Table 2 Track start probability of two algorithms (%)

[0086]

Claims

1. An improved track initiation method based on modified Hough transform, characterized in that: include: Step S100: Extract the first measurement point d from the measurement set D i =(x i ,y i , Tab i ), then extract a Tab that is larger than the first measurement point Tab i The measurement point of is associated with it and the period of the two measurement points is calculated. Step S200: Determine whether the distance threshold and speed threshold between two measurement points are met; Step S300: Calculate the parameters of the straight line between the two points and store the data results in the track parameter set P; Step S400: Traverse the track parameter set P, accumulate and vote for P = (ρ, θ) in the parameter space, select the peak point, match all measurement points that satisfy d ≤ ε1, and remove the measurement points with discontinuous and isolated time series labels in the track parameter set P through time series filtering, retain the continuous points, and then use threshold selection to retain the track parameters with more than m measurement points in the track parameter set P. Step S500: traverse the measurement data in the track parameter P to calculate its position relationship and determine whether it meets the operation rules. If it meets the rules, it is added to the real track set; otherwise, it is considered a false track and added to the false track set. Step S600: The actual track parameter set stores the starting track, and the track initiation is completed.

2. The method according to claim 1, characterized in that The specific process of step S100 includes: Step S101: Extract the first measurement point d from the measurement set D i =(x i ,y i , Tab i ), where the set of measurements to be measured is D = [D t , Tab t ], including all the measured points, D t Indicates the coordinates of the measurement point (x t ,y t ), Tab t Indicates the measurement cycle number of the measurement point, where Tab = 1, 2, 3, ..., N; Step S102: Extract a Tab that is larger than the first measurement point Tab i The measurement points are associated with the measurement points extracted in step S101. Assume that the selected points are d i =(x i ,y i ) and d j =(x j ,y j ), then the period between the two measurement points is Δt=(Tab j -Tab i )×T, where T is the radar scanning period.

3. The method according to claim 1, characterized in that The specific process of step 200 includes: Step 201: Calculate the distance between two measurement points D is obtained by calculation or estimation init , if R≤D init , indicating that the distance between the two points is within the set distance threshold D init Inside; Step 202: Further determine the velocity between the two points and obtain V by calculation or estimation. min 、V max , according to the principle of kinematics, calculate the speed V=R / Δt, if V min <V <V max , which means that the speed threshold is satisfied between the two measurement points, and the process continues to step S300. Otherwise, the process goes to step S102 and a new point d is extracted. j .

4. The method according to claim 1, wherein The specific process of step S300 includes: Calculate the straight line parameter p between two points according to the two-point straight line formula i =(ρ i ,θ i ), the data results are stored in the track parameter set P, where the track parameter set P = [p i , r t ] is a dynamic linked list, where p t Used to store the linear parameters (ρ t ,θ t );r t The points in the xy plane that satisfy the straight line parameters and the label parameters of the measurement cycle to which they belong (x t ,y t , Tab t ); Initialize the track parameter set P = NULL, set the threshold to τ, and the minimum number of starting points of the track to m.

5. The method according to claim 1, wherein The specific process of step S400 includes: Step S401: Traverse the track parameter set P, accumulate and vote for P = (ρ, θ) in the parameter space, and find small areas where more points fall. The center points of these small areas are the peak points. Step S402: The peak point corresponds to a straight line in the xy plane, matching all lines that satisfy d≤ε l The measurement point, where d is the shortest distance from the measurement point to the straight line, ε l A preset threshold for permissible; Step S403: By using time series filtering, the measurement points with discontinuous and isolated time series labels in the track parameter set P are removed, and the continuous points are retained. t Update to the retained consecutive points; Step S404: retaining the track parameters with the number of measurement points greater than m in the track parameter set P through threshold selection.

6. The method according to claim 1, characterized in that The specific process of step S500 includes: Step S501: Based on the measurement cycle number Tab, the measurement data in the track parameter P can be traversed to calculate their positional relationships and determine whether they meet the operating rules. Based on the time series parameter Tab in the track data set P, each track point is sorted in descending order according to Tab = 1, 2, 3, ..., N to construct a time series coordinate matrix: Using the measurement points d of the first two measurement cycles k-1 (x k-1 ,y k-1 ) and d k (x k ,y k ), calculate the velocity component, where d k is a measurement point in the kth cycle: Step S502: Predict the position of the k+1th cycle based on the current speed: The predicted position and the actual measurement point d k+1 Compare the position residuals and calculate the residuals: Step S503: Determine whether it satisfies the dynamic consistency and velocity smoothness. The dynamic consistency requires the residual ε≤ε th , where ε th is the maximum allowed prediction error; velocity smoothness requires that the velocity change rate (i.e., acceleration) of adjacent cycles does not exceed the threshold: where a max is the target maximum acceleration. Step S504: If all periodic measurement points meet the above conditions, the track is retained, duplicate points are removed, and it is added to the real track set. If any period does not meet the motion law, the track is moved to the false track set.

Images