Real-time Moving Object Video Tracking using Support Vector Machines for Visual Servo Application

摘要

Visual servo refers to a process of moving a robot to fulfill certain specific purpose by using visual feedback. For a 2 DOF platform such as pan-tilt camera, the main purpose is to maintain the target object position in the camera field of view, specifically in the center of the image plane. This problem of automatically tracking a moving target has become a research topic worthy of further investigation because of the increasing use in video surveillance and security system. Nonetheless, the success of this problem in real time condition lies primarily on the performance of the visual feedback system used. This paper addresses the visual tracking control of an unknown target that may move arbitrarily in the scene. We proposed a system consist of a SONY FCB-EV7520 camera placed on pan-tilt manipulator platform, and the so-called structured output tracking with kernel (STRUCK) to track a moving target based on the consecutive image acquired. For visual servo control, a simple Proportional-Integral-Derivative controller is used for reducing the system error. Simulations on different conditions have verified the performance of the proposed system.