Research On Target Tracking And Target Contour Extraction In Infrared Image Sequence

Infrared imaging technology has been widely used in military, medical, transportation, monitoring and other fields. Online tracking infrared image target and locating target contour becomes one of the hot research topics in the research field of infrared image technology. According to the characteristics of infrared images, this thesis is accomplished by studying the infrared image target tracking technology based on TLD(Tracking-Learning-Detection) algorithm and infrared image target contour extraction technology based on active contour model.Firstly, the characteristics of infrared images is analysed according to the imaging mechanism, and the histogram model and statistical distribution model of infrared image data are studied. Based on that, a hybrid image enhancement algorithm is proposed to eliminate the drawbacks of low contrast and blurred edge of infrared image, so that the quality of infrared image is improved.Secondly, on the basis of analyzing the principle and process of TLD target tracking method, an improved TLD target tracking method is proposed, which is suitable for infrared image. The improved TLD method integrates scale changeable Meanshift algorithm and Kalman filter into the original TLD method. Experimental results show that the improved TLD algorithm improves the tracking accuracy and speed compared with the original TLD method.Thirdly, the target contour extraction results of several well-known snakes are analyzed. To overcome their shortcomigs, a weighted potential energy model based on distance and a minimum variance model is proposed. Both of these two methods are insensitive to initial contour and noises, and have less computational cost.Finally, by combining the hybrid image enhancement algorithm, the improved TLD target tracking method and the proposed target contour extraction method, an infrared image target tracking and contour extraction system is designed, and is verified in a real scene.