Infrared Dim Small Target Detection In Heavy Clutter Background

Automatic target recognition (ATR) has been researched in several scientific and engineering disciplines. The infrared (IR) image target identification system possesses night vision, anti-hidden capability as well as the mist penetrating power. The system can process high-resolution images in real time, thereby having autonomous precision strike capability of offensive weapons. Moreover, the infrared search and track surveillance system becomes particularly crucial over a large expanse of the sea. The entire route of aircraft guidance consists of auto target identification, tracking, and positioning calculating. This paper discusses the small low observable target detection in cluttered infrared images, which is of interest in many applications such as ocean surveillance (e.g. oil spills), search (e.g. speedboats, dinghies) and rescue (e.g. swimmers), remote sensing and floating mine detection, etc.The aim of the present study is to reduce the false alarm rate of a combined visual and IR surveillance CCD camera system to an acceptable level when the objects are rather dim in relatively dark sea surface backgrounds and the small targets hard to be distinguished from noise and clutters.An adaptive method of dim small target detection in infrared images with a complex background will be investigated in this paper. Firstly, we make a detailed analysis of the characteristics of the background, the target, and the noise in the gray intensity, space and frequency domain of images. Secondly, the modified top-hat transformation using interrelated structuring elements is adopted to adaptively detect the darker and the brighter targets and greatly suppress the cluttered background. Lateral pattern inhibition enhances local contrast ratio and identifies the targets of interest simultaneously. Thirdly, the automatic threshold is used to enhance real dim targets in the cluttered background. Finally, the simulation based on the proposed algorithm is carried out. The results prove that the algorithm is effective and valid.