CN-122017784-A - Radar signal trace point condensation method based on GPU and sparse table

CN122017784ACN 122017784 ACN122017784 ACN 122017784ACN-122017784-A

Abstract

The invention particularly relates to a radar signal trace condensation method based on a GPU and a sparse table, which comprises the steps of running the method on the GPU, firstly rearranging the trace according to a distance ascending order, extracting the distance and the signal to noise ratio in parallel, then constructing the sparse table, carrying out parallel processing among columns, carrying out serial processing among rows, recording the maximum signal to noise ratio in each target interval taking the current trace as the start, then adopting a dichotomy to search left and right boundary indexes of each trace in parallel according to a distance threshold, dividing the left and right intervals of the trace into two longest subintervals with the length of 2 to the power, and searching the corresponding positions of each subinterval in the sparse table, thereby obtaining the maximum signal to noise ratio of each of the left and right boundaries. And finally, comparing the maximum signal-to-noise ratio of each of the left and right boundaries with the signal-to-noise ratio of the trace point, and judging the effectiveness of the trace point. The invention can efficiently distinguish effective points for scenes with larger number of points, and avoids false alarms possibly occurring by adopting a greedy algorithm.

Inventors

WU QIFAN
YANG LEI
LUO DINGLI

Assignees

西安电子工程研究所

Dates

Publication Date: 20260512
Application Date: 20260320

Claims (9)

1. The radar signal trace condensation method based on the GPU and the sparse table is characterized by comprising the following steps of: Acquiring a point trace set of radar signals, wherein the point trace set comprises the distance and the signal-to-noise ratio of each point trace; Ordering the trace point data according to the distance based on the GPU to obtain a distance sequence and a signal-to-noise ratio sequence; Based on the distance sequence, the left boundary index and the right boundary index of each point trace are extracted in parallel by using the GPU, and a left distance neighborhood region and a right distance neighborhood region corresponding to each point trace are obtained; Performing layer-by-layer recursion on the signal-to-noise ratio of each trace in the signal-to-noise ratio sequence in a mode of increasing the interval length, and constructing a signal-to-noise ratio sparse table; and obtaining the maximum signal-to-noise ratio of each trace in the corresponding left-side distance neighborhood region and right-side distance neighborhood region by utilizing the signal-to-noise ratio sparse table to obtain a query result, and determining the effective trace according to the query result and the signal-to-noise ratio of each trace.
2. The method of claim 1, wherein the ordering the trace data by distance based on the GPU results in a distance sequence and a signal-to-noise ratio sequence, comprising: Taking the distance of each trace as a sequencing basis and the signal-to-noise ratio of each trace as accompanying data; based on the distance of each trace, performing parallel ascending sort by using the GPU to obtain a distance sequence arranged according to the distance; And synchronously adjusting the signal-to-noise ratio of each trace according to the sequence of the distance of each trace to obtain a signal-to-noise ratio sequence corresponding to the distance sequence.
3. The method according to claim 1, wherein the extracting, based on the distance sequence, the left boundary index and the right boundary index of each trace in parallel by using the GPU, to obtain a left distance neighborhood region and a right distance neighborhood region corresponding to each trace includes: Distributing a GPU thread to each trace in the distance sequence; Performing a binary search algorithm in parallel by using the distributed GPU threads, searching the positions of the furthest left and right tracks with the distance difference from the current track within a preset threshold range, and respectively determining indexes of the positions as a left boundary index and a right boundary index; and respectively determining a left distance neighborhood zone and a right distance neighborhood zone corresponding to each trace point according to the left boundary index and the right boundary index.
4. The method of claim 1, wherein the step-by-step recursion of the signal-to-noise ratios of each trace in the signal-to-noise ratio sequence is performed in a manner of increasing interval length, and constructing a signal-to-noise ratio sparse table comprises: Directly writing the signal-to-noise ratio of each trace in the signal-to-noise ratio sequence into an initial layer of a sparse table, wherein each table entry corresponds to the signal-to-noise ratio of one trace; For each table item of each layer, inquiring the maximum signal-to-noise ratio of two adjacent subintervals of the previous layer, and taking a larger value of the maximum signal-to-noise ratio as the table item of the current layer, wherein the interval length of each layer is twice the interval length of the previous layer; And completing data construction of all layers of the sparse table through layer-by-layer recursive computation, wherein serial computation is utilized among the layers, and parallel computation is utilized inside each layer.
5. The method of claim 1, wherein the obtaining, by using a sparse signal-to-noise ratio table, a maximum signal-to-noise ratio of each trace in a corresponding left-side distance neighborhood region and right-side distance neighborhood region, respectively, to obtain a query result includes: constructing an index pair table based on the trace index in the trace data; Dividing a left distance neighborhood region and a right distance neighborhood region into two subregions respectively, and acquiring a region level in an index pair table according to the length of the subregions; acquiring the maximum value of the signal to noise ratio in each subinterval in a signal to noise ratio sparse table according to the interval level; And determining the larger signal-to-noise ratio in the signal-to-noise ratio maximum values corresponding to the two subintervals as the maximum signal-to-noise ratio in the distance neighborhood interval.
6. The method of claim 1, wherein determining the effective trace based on the query result and the signal-to-noise ratio of the trace comprises: and when the maximum signal-to-noise ratio in the left-side distance neighborhood region and the right-side distance neighborhood region are both larger than the signal-to-noise ratio of the current trace, determining the current trace as an effective trace.
7. The method of claim 5, wherein the index pair table comprises: Wherein, the Representation of Is used for the indexing of the elements in (c), starting with 0; Representing the trace index in the trace data.
8. A computer readable storage medium, characterized in that the computer readable storage medium comprises a stored executable program, wherein the executable program when run controls a device in which the storage medium is located to perform the method of any one of claims 1 to 7.
9. A computer program product comprising a computer program which, when executed by a processor, implements the method according to any one of claims 1 to 7.

Description

Radar signal trace point condensation method based on GPU and sparse table Technical Field The invention relates to the technical field of radar signal processing, in particular to a radar signal trace point condensation method based on a GPU and a sparse table. Background The radar signal point trace aggregation algorithm realized based on DSP or CPU is usually a greedy algorithm for the point trace number N, and the algorithm time complexity is as follows. As the trace points increase, the algorithm time consuming will increase significantly, which will result in overall signal processing in a state of waiting for hysteresis. Also, at high data rates, further compression of the number of traces that can be processed is required. In addition, if the current trace falls within the right boundary of the distance from another trace and the signal to noise ratio of the trace is low in the processing process, the trace may be marked as an invalid trace, and a false alarm occurs. It should be noted that the information disclosed in the above background section is only for enhancing understanding of the background of the invention and thus may include information that does not form the prior art that is already known to those of ordinary skill in the art. Disclosure of Invention The invention provides a radar signal trace point condensation method based on a GPU and a sparse table, a computer readable storage medium and a computer program product, which can effectively overcome the defects in the prior art. Other features and advantages of the invention will be apparent from the following detailed description, or may be learned by the practice of the invention. According to a first aspect of the present invention, there is provided a method for condensing radar signal traces based on a GPU and a sparse table, the method comprising: Acquiring a point trace set of radar signals, wherein the point trace set comprises the distance and the signal-to-noise ratio of each point trace; Ordering the trace point data according to the distance based on the GPU to obtain a distance sequence and a signal-to-noise ratio sequence; Based on the distance sequence, the left boundary index and the right boundary index of each point trace are extracted in parallel by using the GPU, and a left distance neighborhood region and a right distance neighborhood region corresponding to each point trace are obtained; Performing layer-by-layer recursion on the signal-to-noise ratio of each trace in the signal-to-noise ratio sequence in a mode of increasing the interval length, and constructing a signal-to-noise ratio sparse table; and obtaining the maximum signal-to-noise ratio of each trace in the corresponding left-side distance neighborhood region and right-side distance neighborhood region by utilizing the signal-to-noise ratio sparse table to obtain a query result, and determining the effective trace according to the query result and the signal-to-noise ratio of each trace. In some exemplary embodiments, the sorting the trace data according to the distance based on the GPU, to obtain a distance sequence and a signal-to-noise ratio sequence, includes: Taking the distance of each trace as a sequencing basis and the signal-to-noise ratio of each trace as accompanying data; based on the distance of each trace, performing parallel ascending sort by using the GPU to obtain a distance sequence arranged according to the distance; And synchronously adjusting the signal-to-noise ratio of each trace according to the sequence of the distance of each trace to obtain a signal-to-noise ratio sequence corresponding to the distance sequence. In some exemplary embodiments, the extracting, based on the distance sequence, the left boundary index and the right boundary index of each trace in parallel by using the GPU, to obtain a left distance neighborhood region and a right distance neighborhood region corresponding to each trace includes: Distributing a GPU thread to each trace in the distance sequence; Performing a binary search algorithm in parallel by using the distributed GPU threads, searching the positions of the furthest left and right tracks with the distance difference from the current track within a preset threshold range, and respectively determining indexes of the positions as a left boundary index and a right boundary index; and respectively determining a left distance neighborhood zone and a right distance neighborhood zone corresponding to each trace point according to the left boundary index and the right boundary index. In some exemplary embodiments, the step-by-step recursion is performed on the signal-to-noise ratio of each trace in the signal-to-noise ratio sequence in a manner of increasing the interval length, and a signal-to-noise ratio sparse table is constructed, including: Directly writing the signal-to-noise ratio of each trace in the signal-to-noise ratio sequence into an initial layer of a sparse table, wherein each table entry