Small surface target detection algorithm based on panoramic vision

摘要

Panoramic vision system has wide prospects in maritime surveillance applications. But it is always lack of target detection technology support. According to the complexity of sea-sky background and the specificity of panoramic images, a novel long-range small target detection algorithm based on sea-sky line extraction is proposed in this paper. Firstly, the reasons that gradient Hough circle transform can't extract sea-sky line from panoramic images successfully are analyzed. Secondly, based on the previous analysis, an improved gradient Hough circle transform algorithm is proposed to detect sea-sky line. Finally, local region complexity of panoramic images is calculated to detect long-range small target in sea-sky line area. Experimental results show its applicability and robustness for panoramic images in different conditions.