Small Infrared Target Detection Algorithm Based On Human Visual Mechanisms

Infrared imaging system has a wide application in military due to its good concealment and insensitivity to the weather. Because the infrared target is remote imaging, the number of the target's pixels are often few and their values are low, sometimes even submerged in the background. Especially when the background is complex, it is hard to detect the dim target, so the infrared dim small target detection in complex background has been a difficult problem which attracts great attention to the scholars. Recently, with people's more understanding about visual information processing, the target detection based on the mechanism of human vision has developed a lot and it has attracted much attention in the infrared target detection.The infrared moving target detection method based on human visual system is researched in this paper. Based on many mechanisms of the human visual system it is proposed to get the image's regions of interest(ROIs) firstly, then the targets can be detected in the ROIs according to the saliency difference between the target and the background. Lastly the targets' track trajectories can be gotten by simulating the mechanism of eye movement. In the extraction of ROIs, the saliency map can be calculated by the contrast of infrared target, after the target is enhanced and the background is suppressed the ROIs can be segmented. In the target detection part, the image is processed in the scale space by Dog filter which has the characteristic of human vision, then the targets can be detected according to the saliency difference between each targets. In the tracking algorithm, the PID algorithm is introduced to track the targets' position, then the Retinex algorithm is used to enhance the local region of the suspicious targets. After segmenting the enhanced local region the target's position can be predicted.In this paper, the proposed algorithm is tested in a variety of backgrounds, including the sky and sea background with signal-to-noise ratio, and it is tested by a variety of performance indicators. By experiments the algorithm proposed in this paper can be applied to a variety of complex backgrounds stability, and it can predict the target well when it is submerged in the background.