Research On The Infrared Dim Target Detection

Infrared video now is widely used in the military, the dim target detection in the infrared image is a key technology in the defence and weapon systems such as infrared search and tracking system, infrared early-warning system, and infrared guidance system. The effective methods for real-time dim target detection enhance the practical range. improve the sensitivity and response time of the infrared system, which is of great importance to increase survival probability. This thesis mainly discusses the approaches of dim target detection in infrared image.Due to the prevalence of low contrast, high amount of noise, and some other issues, the pre-processing method must be imposed to the infrared image. This article firstly studied the characteristics of the dim target, background and noise, then simulated and verified the performances of some classical pre-processing methods. Since the size of the target is very small, and the vast majority of the image is the background, it is meaningful to detect the target by background. On the basis of analysis to the current background prediction methods, a concept of comparability filter is proposed. The comparability filter can effectively strengthen the isolated dim target and suppress the gently background, the SNR and contrast of the infrared image are obviously improved. There are always some false points in the single frame, the article comments the modified movable pipeline filter to detect dim target in the multi-frame images, it can avoid the false caused by noise which at the edge of the pipeline. The movement of the target has the regularity and continuity, so the pipeline filter was combined with moving prediction based on these characteristics, after further improvements to the pipeline filter, the method not only improves the accuracy but also can avoid the loss when target is occluded. We use extended state observer to predict the target's movement, and compared it with kalman filter by simulation. The experimental results showed that the proposed target detection algorithms are effective, it can improve the robustness of the system.