Long-distance Front Infrared Target Tracking Method Study

In this thesis, forward looking infrared (FLIR) image sequences are used as the object of research. The basic mechanisms of the infrared imaging and target feature have been analyzed, then a lot of researches on the infrared image preprocessing, target detection, multiple target tracking theory and the algorithm of data association has been done. Lastly joint probabilistic data association and multiple hypothesis tracking were used in infrared image sequences and the tracking of multiple targets was implemented.The mainly researches and innovations as follows:1. After analysis of the infrared imaging features, which has low signal to noise, low contrast and less information of target. In order to make target prominent and easy to detect and tracking, median filter and mean filter was used to denoising and histogram equalization technique was adopted to improve contrast.2. Base on the characteristics of infrared image background, which has a large area which has continuous and uniform distribution, and target has high gray scale, not related to background. A lot of researches on image background suppression have been done, and it makes the image background suppressed and target prominent through the simulations of the background suppression algorithms.3. Multi-target tracking theory was researched, include target tracking formation and termination, design of track gate, data association and tracking maintenance. Focus on kalman filtering and data association. It proved that it can implement target tracking stably in clutter environment through the theoretic simulation on target tracking based on data association and analysis of result.4. On the base of algorithm research before, a kind of multi-target tracking in infrared image based on joint probabilistic data association algorithm is used, According to the characteristics of infrared image and the environment of tracking, multi-target tracking’in infrared image sequences based on joint probabilistic data association and multiple hypothesis tracking is used. It can track multiple targets in infrared image sequences stably through the algorithm.