A Research Of Moving Target Tracking System With Dual FOVs

In recent years, with the price down of camera, the enhancement of computer’s performance and the improvement of network speed, the video surveillance technology has been developed rapidly. However, usually today’s widely used intelligent monitoring system can only be used in narrow backgrounds and cannot adapt to wide backgrounds.Thus we designed and implement a moving target tracking system with dual FOVs. This system can monitor a relatively wide background and capture the image with details of the moving object. The system uses two cameras. One is a camera with wide field of view, the other is a camera with narrow field of view. The wide FOV camera does not move and keeps observing the same place. The narrow FOV camera tracks the object with the help of the motor, and records the details of the target. We show the FOVs’range in a specific scene in this paper.The PC captures images from the wide FOV camera and analyze these images to get the pixel coordinates of the moving object. It converts the pixel coordinates to the horizontal and vertical angle for the narrow FOV camera to rotate and then send the corresponding commands to the motor. Our main work to achieve the system is: (1)We derived a coordinate conversion algorithm based on perspective transformation, did the error analysis and did experiments on it. (2)The algorithm to detect the moving object is Gaussian Mixture-based Background/Foreground Segmentation Algorithm. And we does some morphological processing on the result as well as extracts connected components to locate all moving objects. Then we set up rules to choose one of them to trace. (3)In order to make this tracking system more accurate and stable, we uses Kalman Filter and coordinate conversion to predict the motion path of the object. (4) We wrote the system’s software in Lab VIEW and C++.We experimented this system to track a walking man and obtained good results. The system can locate the target correctly in spite of the interference from some other moving objects. Even when the narrow FOV camera lose the target because of it moves quickly, the camera can capture the target later.The position of two cameras and the angle of the wide FOV camera are quite free, so the system can apply to some scenes that are inconvenient to set up cameras.