Multi-objective Trajectory Tracking Algorithm And Dsp Implementation

Under modern war conditions, it's important to detect seriously threatening martial objects betimes. First, image dynamical the object at a high speed from a long distance away. Then analysis frames of the image and gain the information that is essential for recognition and tracking. Image recognition and tracking system of moving objects is a process of detecting-recognition-tracking, based on computer vision technology and image sequence computing. Finally the object's trace is tracked. And the system should response in time in an abnormal case.A lot of research has been done in this field home and abroad, but it's still difficult to find a universal arithmetic, especially in the multi-target case. Most existing arithmetic is not optimum and has to be optimized in special chips.This paper concentrates on tracking of multi-moving-target. Based on the current research, it presents a new arithmetic and validates by program. In aspect of target capturing, we improce the algorithm of frame difference and background subtraction. For target recognizing, we introduce histogram statistical recognition and image characters matching, which obviously reduces the calculation and improves the speed. Also focusing on target tracking, Kalman motion estimation is optimized. As for the multi-target conditions, based on principle of motion continuum and trace consistence, a algorithm based on motion feature maxing-likely estimate is presented, which has been validated to have satisfied performance and relatively low computation.