Research On Track-Before-Detection Algorithm For Infrared Target Detection

The detection of dim small targets in the infrared images has been the subject of intense investigation for not more than two decades. If you want to get in on the ground floor in modern high-tech war, the target must be found early and detected in a long distance. The basic problem inherent to extent the detection range is the detection of small, low observable, moving targets in images and subsequent estimation of target trajectories. The small spatial extent of the targets limits the information content of targets, and the signal-to-noise ratio is sufficiently low that detection cannot be met by an analysis of a single image frame. This dissertation address the problem of designing new efficient and effective image sequence processing schemes that will successfully detect small targets with very signal-to-noise ratio(SNR). It is called the track-before-detection algorithms(TBD). Track-before-detection algorithms base on the idle model of background and targets, but the actual one is far away from it, so we must use the preprocessing to shorten the distance. The preprocessing is important to the whole detection .it can suppress the background clutter, estimate the background image signal and improve SNR. In this article, we recommend three prewhitening algorithms, one is high-pass filter, and the others are adaptive filter and morphology filter. Each way has its own advantage and disadvantage; you can choose the right one by the actual scene and the expectation.Dynamic programming (DP) and multiple hypotheses testing (MHST) are the representatives of TBD algorithms. The DP algorithm identifies targets by performing the equivalent of an exhaustive search all possible target paths. It treats the small target detection problem as finding the trajectory with maximum grey accumulation. The MHST algorithm is used to prune a tree-structured list of candidate trajectory segment at each pixel, the trajectory contain an object is accepted. The emluator show us that using this two methods can detect small target accurately, but they take system unendurable memory and computation, it must be improved to use in fact.At the last chapter, an Update algorithm basing on DP is proposed here. It combines the multistage thresholds of MHT with DP theory. This action hold the accuracy of the DP algorithm, at the same time, it save the memory and computation of system.