Research On Some Key Technologies For Real-Time Visual Tracking Using Infrared LED Markers

Visual tracking, which is a key technology of virtual reality, human-computer interaction, and visual surveillance, has tremendous prospects in application. Because of the occlusion and complicated image processing, the real-time visual tracking, which is the current research focus, is difficult to improve. Visual tracking based on the infrared LED markers that is obvious and easy to identify can best reduce background change for feature detection. This thesis works on a number of key technologies of real-time visual tracking based on infrared LED.Three different ways, including contour extraction, regional segmentation, and histogram, were investigated in this thesis. Through a comparative analysis on stability, time-consuming and accuracy, the contour extraction was considered to be the optimal approach and used in our studies. Kalman Filtering was used in this tracking application. Analyzed two way of Kalman filtering: the image coordinate system and the world coordinate system Kalman filter, found the suitable one. We analyzed the method how the multi-camera feature information synthesize the world coordinate information of the markers. The experimental platform and a prototype software system were developed after the analysis of system performance.Infrared LED feature is very obvious, and the impact of changes in the background is small in feature extraction. Also, Kalman filtering effectively reduced the search area of feature extraction, making feature extraction more rapid. The forecast results can also be used in tracking failure instance and this improves the robustness. Using multi-constraint into the rapid synthesis of infrared LED markers information in the world coordinates. All of these features enhance the performance of infrared-based real-time visual tracking.