Research On Active Vision Tracking System Based On Image Information

Computer vision refers to a kind of simulation to biological vision by using computers and related equipments. Its main task is to gain three-dimensional information from a certain scene through operating collected pictures or videos. And active vision is a hotspot in computer vision, it is combined with image processing, pattern recognition, artificial inteligence and other advanced technology, it can shift the visual angle of camera actively and intelligently, and has a greater range and flexibility compared with passive vision. So active vision can be widely used in modern military, medical system, intelligent surveillance, robotic vision system and other fields.The article mainly studied the object detection and tracking of active visual system in the application background of video surveillance.Firstly, the article established the framework of active visual system, and system consisted of camera amaging system, communication system and PTZ system, and laid the foundation of subsequent data processing and research work of active vision tracking.Secondly, as for the moving object detection, the article analyzed the advantage and disadvantage of three common methods including the frame subtraction method, background subtraction method, optical flow method, and the detection algorithm based on three frames subtraction and background subtraction is presented. Through the experiment contrasts, the algorithm could detect the real-time object in the video images effectively.Finally, as for the moving object tracking, the article selected the tracking algorithm for the fusion of EKF and RBF neural network, and used extended Kalman filter to do the optimal estimation for the next state based on the previous sequence of state, and predicted the possible place of object. We could get the trajectory in the range of the wide-angle-lens camera view, and achieve the active vision tracking of object. The experimental results showed that the system has the advantage of real-time performance and robustness.