Research On Tracking Before Detection Algorithms Of Dim-Small Targets Based On MeMber Filter

Under low signal-to-noise ratio(SNR), the detection and tracking of infrared dim-small targets is the key technology of photoelectric detection and tracking system. With the development of modern science, the performance of vehicle maneuverability and anti stealth is higher, which puts forward the strict request to the detection and tracking. The random set theory is applied to the field of dim target detection and tracking. The combination of multi-target tracking algorithm with TBD technology to implement detection and tracking of unknown number of dim target with low SNR, enriches the content of random set theory and highlights the advantages of TBD algorithm. This thesis focuses on the research of Me MBer-TBD algorithm for detecting and tracking dim-small targets, then some new improved algorithms is proposed and implemented.First of all, Me MBer filter based on random set and the model of infrared dim-small targets with TBD algorithm are introducted in detail. Then the thesis emphasizes the algorithm principle of Me MBer-TBD and its implementation based on gaussian particle. In the complex scene with cross and adjacent multi-target, the traditional merging algorithm usually caused the combining error and resulted in missing target information and poor stability, so the thesis puts forward an improved merging algorithm which based on label. Simulation experiments with multiple scenarios show that the proposed algorithm can realize stabilized detection and tracking perfectly, especially under low SNR, it still can perform the target merged accurately.Then, for the problem of high storage and computation complexity of traditional TBD algorithm, a threshold-measurement TBD algorithm is proposed according to the sensor's point spread function and applied to the infrared dim target detection combined with Me MBer filter. Instead of raw image data, index set is used to be measurement set and detection probability is used to calculate likelihood function. Simulation results show that the algorithm is effective. Meanwhile, because of the available thresholded sensor, the algorithm has a good application prospect in the actual application.Finally, for the problem of false alarm and serious missing detection with the traditional image preprocessing algorithm, an improved Robinson Guard background suppression algorithm is proposed. For the instability of the detection and tracking, a tracking algorithm based on SCK-MB-TBD is put forward. Experiments show that the improved Robinson Guard has the better clutter suppression effection and the stronger ability to keep target's information. At the same time, SCK-MB-TBD could guarantee the tracking real-time and numerical stability. Because the number of cubature points is related to filter dimension, the performance will be improved with the increase of state dimension to a certain extent.