Research Of Intelligent Human Target Tracking Technology Based On Infrared Video

Infrared images are obtained by the infrared sensor with its thermal imagingperformance, which are only dependent on temperature and heat radiation of the objects.Relative to the visible imaging, the infrared imaging system can obtain targets even inthe smoke or under the condition of insufficient light with its unique night visionperformance. Target tracking based on infrared image has played an important role inmilitary area such as infrared guidance and reconnaissance as well as the civilian areasuch as video surveillance and driver assistance. Tracking of infrared pedestrian target isa hot topic and challenging research in target tracking field as human target is the mostimportant and most active factor in the environment. However, this task is full ofdifficulties and challenges because of the non-rigid human body and the infraredimage’s own shortcomings.This topic is the study of the intelligent human target tracking technology based oninfrared video. Our main work is to develop feasible and stable infrared human trackingalgorithm which also has strong robustness, then, integrate the tracking algorithm intothe control of the pan-tilt and realize the function of the system to track the pedestrianautomatically.On the study of the algorithm, first, we make thorough analysis of thecharacteristics and difficulties of the infrared human target tracking and chose MeanShift algorithm as the theoretical basis considering the real-time requirement of thesystem, and then, we deeply analysis the principle of the Mean Shift algorithm also thesetting of the parameters. Second, we propose a method of infrared pedestrian targettracking based on Mean Shift with boundary constraint and a method of target trackingusing Mean Shift combined with kalman filtering, the former uses the gradient as wellas the brightness information of infrared image processed by anisotropic diffusion toobtain the target boundary and uses the strategy of Mean Shift algorithm to realize thehuman tracking, the latter uses kalman filter to predict the target’s location and sets theprediction position as the initial point of the Mean Shift iteration to resolve the rapidmotion problem, and also judges the occlusion by Bhattacharyya coefficient, then,proposes the tracking strategy under the condition of occlusion, moreover we proposesthe tracking strategy to cope with the problem of target’s long time hidden, at last, therobustness of the algorithm is improved through the above improvement. On the study of the tracking system, first, we introduce the construction of thehardware platform for the system, which is consists with the infrared thermal imager(ThermoVision A40), two-dimensional combined pan-tilt and PC and then give theimplementation process of the system and describe the function of each module. Second,we propose the control strategy of the pan-tilt, in which we divide the image area intothree aspects and use the human target’s area to guide the movement of the horizontaland vertical frame of the pan-tilt to keep the target in the camera’s sight. At last, weintroduce the software design of the system and then we verify the function of thesystem through the experiments.