A method for distinguishing a nearby target based on local peak-valley alternating search

Abstract

The application discloses a near target distinguishing method based on local peak-valley alternately searching, and is suitable for the case that near targets are condensed into a single target. The application can split the point traces into multiple targets to realize the distinguishing of the condensed near targets and improve the distinguishing performance of the near targets by alternately searching the local peak-valley points and splitting the data into multiple targets at the local valley points.

Description

A method for distinguishing nearby targets based on alternating search of local peaks and valleys Technical Field

[0001] This invention relates to the field of radar target detection, specifically to a method for distinguishing nearby targets based on alternating search of local peaks and valleys, applicable to situations where the detection points of nearby targets are aggregated into a single target. Background Technology

[0002] In radar target detection, the echo of a single target, after constant false alarm rate (CFAR) decision, is typically composed of multiple detection points. Therefore, a crucial step is to agglomerate these multiple detection points into a single target. However, situations inevitably arise where multiple targets are very close together, such as ships chasing each other at sea. In such cases, waves or spray generated by the ships also create detection points, causing the detection points of nearby targets to merge together. Consequently, commonly used point aggregation methods based on undirected connected regions agglomerate nearby targets into a single target. Therefore, studying the differentiation of nearby targets is of great significance for distinguishing near targets at sea. Summary of the Invention

[0003] In view of this, the present invention proposes a method for distinguishing nearby targets based on alternating peak and valley search. After constant false alarm rate decision and point clustering, the column sum of the detection point data clustered into a single target is calculated and peak and valley alternating search is performed based on the column sum. Then, the detection point is split into multiple targets at the valley point position, which can distinguish nearby targets and improve the resolution performance of nearby targets in the case of multiple targets.

[0004] The objective of this invention is achieved as follows:

[0005] A method for distinguishing nearby targets based on alternating local peak and valley values ​​includes the following steps:

[0006] Step 1: After constant false alarm rate (CFAR) decision and point clustering, take the data s corresponding to the detection point of the current target on the range-Doppler map. m，n Calculate s along the Doppler dimension m m，n The maximum value, and for s m，n Summation, denoted as the maximum value of the column x. n and column and y n Find the maximum value x in two dimensions. max :

[0007]

[0008]

[0009]

[0010] Where n = N0, N0+1, ..., N0+N-1, m = 1, 2, ..., M, N0 and N are the first range cell and the number of range cells occupied by the detection point of the current target, respectively, and M is the number of Doppler cells. When the detection point belongs to the current target and exceeds the constant false alarm rate (CFAR) threshold, s... m，n Retain the values ​​on the distance-Doppler plot; otherwise, s m，n =0;

[0011] Step 2: Set the local peak point filtering threshold factor T1, the local valley point filtering threshold factor T2, and the neighborhood size R. Initialize the search state to search for local peak points, search index n = N0, and the local valley point sequence φ to be empty.

[0012] Step 3: If the current search state is searching for local peak points, proceed to step 4; if the current search state is searching for local valley points, proceed to step 6.

[0013] Step 4, determine x n Is it a local peak point? If so, proceed to step 5; otherwise, proceed to step 8.

[0014] Step 5: Record the column sum corresponding to the currently found local peak point, i.e., y. peak =y n Then change the search status to search for local valley points, and proceed to step 8;

[0015] Step 6, determine x n Is it a local valley point? If so, proceed to step 7; otherwise, proceed to step 8.

[0016] Step 7: Add the current index n to the local valley point sequence φ, record the number of elements P in the sequence φ, change the search state to search for local peak points, and then go to step 8;

[0017] Step 8, update n = n + 1, if n > N n +N-1 indicates that the distance cell search is complete, proceed to step 9; otherwise, proceed to step 3 to search for the next distance cell.

[0018] Step 9: If the search state is searching for local peak points and the number of elements P in the local valley point sequence φ is greater than 0, then remove the last element in the sequence φ and update the number of elements P = P-1 in the sequence φ.

[0019] Step 10: If the number of elements P in sequence φ is greater than 0, proceed to step 11; otherwise, it means that no local valley point has been found, and the method ends.

[0020] Step 11, divide the data s according to the elements in sequence φ. m，n Divide into P+1 objectives, i.e.

[0021] ψ1={s m，n ,N0≤n≤φ1}

[0022] ψ p ={s m，n , φ p-1 <n≤φ p}, p = 2, ..., P

[0023] ψ P+1 ={s m，n , φ P <n≤N0+N-1}

[0024] Wherein, the subscript φ indicates the nth element in the sequence φ;

[0025] Complete the differentiation of nearby targets.

[0026] Furthermore, in step 4, determine x n Whether it is a local peak point, the specific method is to determine x. n Does it meet the following two conditions:

[0027] x n >T1x max

[0028] x n >x r , r∈[nR, n+R], r∈[N0, N0+N-1], r≠n.

[0029] Furthermore, in step 6, determine x n Whether it is a local valley point, the specific method is to determine x. n Does it meet the following two conditions:

[0030] x n <T2y peak

[0031] x n <x r , r∈[nR, n+R], r∈[N0, N0+N-1], r≠n.

[0032] The beneficial effects of this invention are as follows:

[0033] 1. This invention has a good ability to distinguish nearby targets. When the detection points of nearby targets are agglomerated into a single target, the detection points can be split into multiple targets, thereby improving the performance of distinguishing nearby targets in the case of multiple targets.

[0034] 2. After constant false alarm rate decision and point aggregation, the present invention calculates the column sum of the detection point data aggregated into a single target and performs peak-valley alternating search based on the column sum. Then, at the valley point position, the detection point is split into multiple targets, which can distinguish nearby targets and improve the resolution performance of nearby targets in the case of multiple targets. Attached Figure Description

[0035] Figure 1 is a flowchart of the present invention.

[0036] Figure 2 is a range-Doppler plot containing three ship targets.

[0037] Figure 3 shows the distribution of detection points for a single target obtained from constant false alarm rate (CFAR) decision and point aggregation.

[0038] Figure 4 shows the distribution of detection points for the three targets obtained by this invention. Detailed Implementation

[0039] As shown in Figure 1, a method for distinguishing nearby targets based on alternating local peak and valley values ​​includes the following steps:

[0040] Step 1: After constant false alarm rate (CFAR) decision and point clustering, take the data s corresponding to the detection point of the current target on the range-Doppler map. m，n Find the maximum value and sum along the Doppler dimension m, and denote them as the column maximum values ​​x. n and column and y n Find the maximum value x in two dimensions. max ,Right now

[0041]

[0042]

[0043]

[0044] Where n = N0, N0+1, ..., N0+N-1, m = 1, 2, ..., M, N0 and N are the first range cell and the number of range cells occupied by the detection point of the current target, respectively, and M is the number of Doppler cells. When the detection point belongs to the current target and exceeds the constant false alarm rate (CFAR) threshold, s... m，n Retain the values ​​on the distance-Doppler plot; otherwise, s m，n It is 0.

[0045] Step 2: Set the local peak point filtering threshold factor T1, the local valley point filtering threshold factor T2, and the neighborhood size R. Initialize the search state to search for local peak points, search index n = N0, and the local valley point sequence φ to be an empty set. In the implementation example, the threshold factor T1 is set to 0.1, the threshold factor T2 is set to 0.3162, and the neighborhood size R is set to 3.

[0046] Step 3: If the current search state is searching for local peak points, proceed to step 4; if the current search state is searching for local valley points, proceed to step 6.

[0047] Step 4, determine x n Is it a local peak point, that is, does it meet the following two conditions:

[0048] x n >T1x max

[0049] x n >x r , r∈[nR, n+R], r∈[N0, N0+N-1], r≠n

[0050] If the condition is met, it is a local peak point, proceed to step 5; otherwise, proceed to step 8.

[0051] Step 5: Record the column sum corresponding to the currently found local peak point, i.e., y. peak =y n Then change the search status to search for local valley points, and proceed to step 8.

[0052] Step 6, determine x n Is it a local valley point, that is, does it meet the following two conditions:

[0053] x n <T2y peak

[0054] x n <x r , r∈[nR, n+R], r∈[N0, N0+N-1], r≠n

[0055] If the condition is met, it is a local valley point, proceed to step 7; otherwise, proceed to step 8.

[0056] Step 7: Add the current index n to the local valley point sequence φ, record the number of elements P in the sequence φ, change the search state to search for local peak points, and then proceed to step 8.

[0057] Step 8: Update n = n + 1. If n > N0 + N - 1, it means the distance cell search is complete, and proceed to step 9; otherwise, proceed to step 3 to search for the next distance cell.

[0058] Step 9: If the search state is searching for local peak points and the number of elements P in the local valley point sequence φ is greater than 0, then remove the last element in the sequence φ and update the number of elements P in the sequence φ to P = P - 1.

[0059] Step 10: If the number of elements P in sequence φ is greater than 0, proceed to step 11; otherwise, it means that no local valley point has been found, and the method ends.

[0060] Step 11, divide the data s according to the elements in sequence φ. m，n Divide into P+1 objectives, i.e.

[0061] ψ1={s m，n ,N0≤n≤φ1}

[0062] ψ p ={s m，n , φ p-1 <n≤φ p}, p = 2, ..., P

[0063] ψ P+1 ={s m，n , φ P <n≤N0+N-1}

[0064] In the embodiment, the range-Doppler map containing three vessel targets, as shown in Figure 2, was processed using constant false alarm rate (CFAR) decision and target aggregation, resulting in the image shown in Figure 3. The detection targets of the three targets were aggregated into a single target. The result of this method is shown in Figure 4, where the detection targets of the three targets were split into three targets, consistent with the actual situation. This demonstrates that this method can distinguish nearby targets aggregated into a single target, exhibiting good resolution performance for nearby targets in multi-target scenarios.

[0065] In summary, this invention, by alternately searching for local peak and valley points and splitting the data into multiple targets at the local valley point locations, can split the trace into multiple targets, thereby distinguishing nearby targets that are clustered together and improving the resolution performance of nearby targets.

Claims

1. A method for distinguishing nearby targets based on alternating search of local peaks and valleys, characterized in that, Includes the following steps: Step 1: After constant false alarm rate (CFAR) decision and point clustering, take the data s corresponding to the detection point of the current target on the range-Doppler map. m，n Calculate s along the Doppler dimension m m,n The maximum value, and for s m，n Summation, denoted as the maximum value of the column x. n and column and y n Find the maximum value x in two dimensions. max : Where n = N0, N0+1, ..., N0+N-1, m = 1, 2, ..., M, N0 and N are the first range cell and the number of range cells occupied by the detection point of the current target, respectively, and M is the number of Doppler cells. When the detection point belongs to the current target and exceeds the constant false alarm rate (CFAR) threshold, s... m，n Retain the values ​​on the distance-Doppler plot; otherwise, s m，n Step 1: Set the local peak point filtering threshold factor T1, the local valley point filtering threshold factor T2, and the neighborhood size R. Initialize the search state to search for local peak points, with the search index n = N0 and the local valley point sequence φ empty. Step 2: If the current search state is searching for local peak points, proceed to step 4; if the current search state is searching for local valley points, proceed to step 6. Step 4: Determine x. n Is it a local peak point? If yes, proceed to step 5; otherwise, proceed to step 8. Step 5: Record the column sum corresponding to the currently found local peak point, i.e., y. peak =y n Then change the search state to search for local valley points, and then proceed to step 8; in step 6, determine x. n Is it a local valley point? If yes, proceed to step 7; otherwise, proceed to step 8. Step 7: Add the current index n to the local valley point sequence φ, record the number of elements P in sequence φ, and change the search state to search for local peak points. Then proceed to step 8. Step 8: Update n = n + 1. If n > N0 + N - 1, the distance cell search is complete; proceed to step 9. Otherwise, proceed to step 3 to search for the next distance cell. Step 9: If the search state is search for local peak points and the number of elements P in the local valley point sequence φ is > 0, remove the last element from sequence φ and update the number of elements P = P - 1. Step 10: If the number of elements P in sequence φ is > 0, proceed to step 11; otherwise, no local valley point was found, and the method ends. Step 11: Calculate the data s based on the elements in sequence φ. m，n Divide into P+1 targets, i.e., ψ1={s m，n , N0≤n≤φ1}ψ p ={s m，n , φ p-1 <n≤φ p }, p = 2, ..., Pψ P+1 ={s m，n φ P <n≤N0+N-1} where the subscript φ indicates the nth element in the sequence φ; to complete the differentiation of nearby targets.

2. The method for distinguishing nearby targets based on alternating local peak and valley searches according to claim 1, characterized in that, In step 4, determine x n Whether it is a local peak point, the specific method is to determine x. n Does it meet the following two conditions: x n >T1x max x n >x r , r∈[nR, n+R], r∈[N0, N0+N-1], r≠n.

3. The method for distinguishing nearby targets based on alternating local peak and valley searches according to claim 1, characterized in that, In step 6, determine x n Whether it is a local valley point, the specific method is to determine x. n Does it meet the following two conditions: x n <T2y peak x n <x r , r∈[nR, n+R], r∈[N0, N0+N-1], r≠n.

Images