反无人机图像导引头远距空中目标探测技术

Taking the low-cost counter-UAV guided munitions as the research object, this study focuses on long-distance UAV target detection, and the design of a local visual saliency-based clustering measurement algorithm. In this algorithm, the idea of local contrast measurement is introduced into the visible light image for the first time, and the detection of the target by measuring the spectral clustering of the local imaging domain is realized. More specifically, it measures the minimum distance between the average spectral value of the local image domain and the spectral values of adjacent pixels. In addition, to solve the multi-scale target problem, a corresponding multi-scale sliding window measurement method is designed. The brief flow of the whole algorithm is as follows: frequency-division median filtering was performed on the original RGB image frame; to better measure the spectral difference, the filtered RGB image was converted to the Lab color space; the sliding window model was used for UAV imaging domain search; the saliency detection method was adopted to measure the spectral difference to obtain a saliency measurement map; finally, a thresholding algorithm was used to obtain the pixel position of the potential UAV target. According to the UAV target imaging conditions, field shooting and artificial synthesis of long-distance UAV image datasets were carried out. The experimental results showed that the algorithm can successfully separate the UAV target from the background under various complex meteorological conditions.