Research On Infrared Image Preprocessing And Dim Targets Detection

Dim and small point targets detection in infrared image is a key technology in 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 and small targets detection of images with low SNR will 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 deals with the infrared image preprocessing technologies and targets detection methods. For image preprocessing, firstly, the motion between images is detected with co-occurrence matrix, then a contour-based approach is employed for image registration to compensate image motion. Secondly, many different image smoothing techniques and background clutter rejection algorithms have been summarized and a comparison is made according to the clutter rejection performance. For targets detection, several classical targets detection algorithms are introduced, based on which an adaptive least median square filter is brought forward to detect dim targets. Experiments show that such filter scheme is superior to time domain median filter and can eliminate most false targets while keeping the true ones. Finally, to detect dim targets in sequence frames, a motive pipeline filter based detection method is proposed, which adopts constant false alarm probability. The experimental results show that this method is capable of detecting multi-targets with 0.5 or more SNR. Besides, this method is simple and easy to implement.