Infrared Small Target Detection Method In Low Snr

IR moving target detection and tracking under complex environment is a key technology problem in IR guidance system and IR alarm system to detect and track the point target in the far distance and the plane target in the near distance. A lot of low altitude objects and the characteristics of point target in high noise and clutter make the detection and tracking even harder. For solving the problem, we use the method of tract before detect(TBD), and preprocess before using TBD to accumulate energy of multiple frames. The goal of preprocessing is to promote SNR(Signal Noise Ratio) and restrain the background of IR images. It gives assistance to improve efficiency of the multiple frames accumulation. By analyzing the merits and demerits of several kinds of preprocessing, such as highpass filtering, median filtering, Tophat transformation in Mathematical Morphology and wiener filtering, the Local Singularity filtering based on improved SUSAN principle has been given, and it has a high power to raise the probability of target detection in single frame. In the TBD detection, dynamic programming(DP) technique was used. DP technique is a high efficiency algorithm to solve combinatorial optimization problems by divided the multidimensional problem into multiple 1-dimension problems. When this technique resorts to accumulate multiple frames, there are some disadvantages. First, the energy would diffuse after multiple frames accumulation and bring agglomeration effect. Second, the amount of calculation and storage level are huge when the SNR is low, especially the number of accumulated frames is large. To solve the problems, two methods have been resorted. Orientation weighting restricts the direction of trajectories by predicted candidate trajectories and shrink ferret window, and reduces the amount of calculation. Truncated sequential probability ratio test(TSPRT) is also a algorithm of TBD, it can improve calculation efficiency by multistage thresholds to truncate tree-structured list of candidate trajectory of low degree of confidence. Combining both DP and TSPRT algorithm, the amount of calculation and storage level are reduced. The update algorithm via a number of simulations by applying the algorithm to both real world and simulated IR image sequences has been demonstrated, and the simulations show that the update algorithm is...