Study And Implementation On Detection And Tracking Dim Small Multi-Targets From Image Sequences

This paper study the detection and tracking of dim small multi-targets from the image sequences.In this thesis, we propose a preprocessing based on background suppressing to suppress the background clutter, estimate the background image signal and improve SNR.This paper address an algorithm of detection and tracking dim small multi-targets based on motion vector estimation. Used the image differencing approach to detect targets, then after the preprocessing and the adaptive double threshold segmentation, the dim small targets and a few high frequency noises can be distinguished effectively. In multi-targets data association, established the search region by motion vector estimation, and then provided a cost function as the data association principle. Finally, give a method to kick out false targets by using the pipeline interval span filter and maintain tracking correctly.At last, this thesis addressed a multi-targets detection and tracking algorithm based on particle filter. Established the multi-targets tracking framework based on particle filter, and then provided methods for the association of particle filter solving joint multi-targets tracking.The above algorithms are carried on emulation experiment with the real infrared image sequences, and are verified the efficient and robust characteristics of algorithms and conclusions from the experimental result.