Investigation On Dim Point Target Detection In Low SNR Infrared Image

The detection and tracking of dim point targets in the optical and infrared images has been the subject of intense investigation for no more than two decades. The optical and infrared sensors are passive sensors, which are valued for their strong survival capacity in battlefields, but their maximum detection range is critical. The basic problem inherent to extent the detection range is the detection of small, low observable targets in images and subsequent estimation of the target trajectories. The difficulties of the detection of dim points lies in the follows: the signal-to-noise ratio is sufficiently low due to the large detection range, no information content of targets signature can be used. Therefore, comparing with other topics in the field of infrared targets detection and tracking, how dim point targets can be robustly detected and tracked under complex backgrounds have become the more realistic and challenging research topics.The problem of dim point targets detection and tracking in low signal-to-noise ratio infrared images has been discussed in this dissertation. A novel thought is provided in this filed. The dim point targets detection algorithm could be divided into two classes, detection based on signal fame and detection based on association, and both are deeply discussed and researched in this dissertation. The main contributions of this dissertation are summarized as follows:(1) As the problem of infrared image preprocessing under complex backgrounds is concerned, the characteristics of dim point target, infrared background and noise during infrared image acquisition is analyzed in this dissertation. Some traditional image preprocessing algorithms are analyzed, aiming at enhancing the low signal-to-noise ratio(SNR) of original images and improving the detection performance. Based on spatial distributions of targets, background and noise, a novel spatial matching filter to raise SNR from original low level is proposed.(2) As the dim point targets detection based on single frame is concerned, a model for the infrared image is constructed. Then a data processing flow for target detection is proposed, which is consist of image preprocessing, background estimation and suppression, adaptive threshold segmentation and targets' statistical characteristics analysis.(3) An improved mathematical morphological filter is proposed. The performance of the traditional mathematical morphology is not as good as expected while it is applied to the dim point targets detection in low signal-to-noise infrared images. To solve the problem, a novel morphology method is presented to reduce background and thus enhance target saliency, through improving the morphological transform and reconstructing the structure element. Finally, a threshold selection algorithm is adopted to extract dim point targets adaptively. In all the stages, diffraction effects of targets and intensity difference of targets with background are taken into account to improve detection performance. Experimental results exhibit the algorithm developed in this dissertation has better and robust performance.(4) As the dim point targets detection based on multi-frame is concerned, the multiple hypothesis tracking algorithm is analyzed. The thought of track before detection and sequential detection is introduced into the multi-frame detection.