Research In Spaceborne Moving Point Target Detection Algorithms For Ir Images

In order to detect the space point target for IR images in real time, this thesis does research in algorithms for point target detection. Performance of the algorithms is verified by MATLAB simulation. Theoretical analysis is carried out to determine that the algorithms in this thesis are fit to the hardware platform based on FPGA.This thesis does research in the traditional detection algorithms and their performance. To reduce the level of the detection false alarm of the traditional detection algorithms, the parallel-processing method is introduced to the new algorithms to process multi-target images and the switch-processing method is applied to the new algorithms to process single-target images. Based on the switch-processing method, edge detection operators are used to reduce the level of the detection false alarm caused by background region edge. New ratio threshold method is brought forward for switch-processing method and edge detection method. Compared to the traditional algorithms, the above new algorithms can reduce the level of detection false alarm rate to half of that of the traditional algorithms while the detection rate of the new algorithms is kept to essentially same level.To fit to the hardware platform based on FPGA, Methods for object track association in image sequences are discussed. For single images, the relationship among SNR and detection rate and false rate is discussed. For sequence images, the relationship between total detection rate and detection rate in single images and the relationship between total false alarm rate and false alarm rate in single images are analyzed. At last, result of theoretical analysis and result of actual test are compared and analyzed.