Detection Of Infrared Weak Targets Based On PSO

The detection of infrared small weak targets and distributed weak targets under complicated background is a difficult and challenging task. With complicated background, usually the small weak targets are only several pixels in size that lack geometrical structure and their gray characteristics are not obvious; simultaneously, the disturbance of clutters makes the target detection even harder. It makes the traditional threshold segmentation based on gray value and edge detection invalid because of the inherent characteristics of infrared imaging, complicated background and the existence of too much background clusters that close to the objects gray value, though some geometrical structure exists in distributed weak targets image. To solve above problems, this dissertation deeply studies the detection for small weak target and distributed weak target and presents some novel algorithms based on Particle Swarm Optimization (PSO).This dissertation firstly introduces the background and development situation of the detection for small weak target and distributed weak target. Then, it elaborates the principles and models of basic PSO and several improved algorithms, and introduces the applied circumstance of PSO in image processing and other fields. An image segmentation model is established based on PSO by combining it with image segmentation issues. Subsequently, a new small weak target detection algorithm is presented, in which suppression of complicated natural background by means of the maximum-minimum filtering and the wavelet inter multiply algorithm is taken into account first, and then a contrast segmentation method based on PSO is utilized to extract the small objects. Finally, the method of infrared distributed weak target detection is studied from the aspect of threshold segmentation. A new algorithm based on PSO and local recursive 2-D maximum entropy segmentation is proposed here, which gets better anti-noise performance by considering simultaneity the distribution of gray information and the spatial neighbor information, but makes the calculating process extraordinary complex and slow. Hence the idea of recursive and PSO is introduced to obtain the threshold vector accurately and fast. The experimental results show that the proposed algorithms make the detection of target effectively.