An Improved BING/NMS Algorithm for Aircraft Detection

摘要

The detection method of the aircraft target should meet the requirements of real-time and precise. Due to the high concealment and clustering of aircraft, the traditional BING algorithm has a lot of false positives and false negatives, which greatly reduces the detection probability of targets. Aiming at the problem, an improved BING/NMS algorithm based on Otsu-HSV color model for aircraft detection is presented in this paper. Firstly, in the case of complex dynamic environment background, the RGB color model is not conducive to image detection of aircraft. Therefore, the reconstruction Otsu-HSV model combing HSV color model with Otsu algorithm for image preprocessing is proposed. Then aiming at the BING algorithm and the NMS algorithm, which will lead to high false detection and miss detection, we propose to improve the detection precision and real-time by using overlap and weighted score. The results demonstrate that the proposed algorithm can improve AP and ABO, reduce false alarm and has a good detection effect.