Divide-and-conquer detection method based on semantic segmentation network

Abstract

The invention discloses a divide-and-conquer detection algorithm based on a semantic segmentation network. According to the method, clutter region segmentation of a distance Doppler domain is realizedby means of a semantic segmentation network, a detection strategy is formulated according to the characteristics of different clutter region main bodies and edges, appropriate detectors and reasonable detection parameters are selected from alternative detector groups to complete divide-and-conquer detection, and false alarms are effectively suppressed while a target is detected to the greatest extent. The technology is of great significance to radar detection of low-altitude small targets. Compared with a detection plane normalization method, the method has the advantages that detection planeconversion is not needed, the calculated amount is smaller, the influence of a parameter estimation result on the detection plane conversion is avoided, and the robustness is higher.

Description

technical field [0001] The invention belongs to the technical field of low-altitude small-target radar detection, and in particular relates to a divide-and-conquer detection method based on a semantic segmentation network. Background technique [0002] Low-altitude small targets are flying targets with a flying altitude below 1,000 meters and a radar cross-sectional area of ​​less than 2 square meters. Low-altitude small targets are closely related to human production and life. According to relevant statistics, biological targets such as birds are the main cause of bird strikes, and man-made targets such as drones are the main tools of terrorist attacks. As a powerful tool for detecting low-altitude small targets, radar is of great significance for early warning and improving public safety. [0003] In the radar detection of low-altitude small targets, due to the unique motion characteristics of the target, it is extremely susceptible to various clutters such as ground vege...

AI Technical Summary

Problems solved by technology

The traditional target detection method is to use a single detector to perform Chinese Constant False Alarm Rate CFAR (CFAR) detection in the range Doppler domain, which cannot take into account the detection performance of different clutter areas, and is prone to false alarms and missed detections

The current processing idea is to identify the clutter area and adopt a differentiated processing strategy. After the clutter area is identified, the idea of ​​distribution indexization is often used, that is, the detection planes containing different clutter areas are normalized to the same parameter Exponential distribution, and then target detection, but this method needs to convert each value of the detection plane, the amount of calculation is huge, and it relies heavily on the parameter estimation results. When the parameter estimation is inaccurate, a uniform detection plane cannot be formed, and the detection performance is sharp. deterioration

Images