Object Location And Motion Analysis Based On Two-Camera Visual System

Two-camera stereo visual system is an important branch of robotic visualtechnologies. It means that two cameras on different locations or one camerafrom dissimilar view angles via translating or rotating shoot the same scene toget the parallax according to which the space position information to the objectcenter point is calculated. The study of two-camera visual platforms includescameras calibration, image processing, stereo matching, three-dimensionalreconstruction and movement analysis. Through the further research oftwo-camera visual technologies, it establishes technical foundations of roboticvision problems such as smart mobile robot navigation and industrial robotsvisual detection and manipulator grabbing.This paper firstly introduces two-camera visual system researchingpurposes, signification and structure of dissertation;then it introduces twomethods of camera calibration, for one thing the two-step method based onRAC was expounded, for another thing a new calibration method based onfuzzied neural network is proposed;and next the methods of image processingtechniques are researched, and the basic operations of image such as filtering,edge detection and image thinning used VC++ programming language areexecuted;after that the experiments and analyses against to object locationsystem are taken, which mainly established a software platform to achieveoperations from the image capturing, processing, camera calibration to objectlocation, and the location precision is analyzed;lastly, the object motionestimation based on Kalman filter is discussed, which has already maturelyapplied in one-step motion estimation. In this paper the method is applied tolong term estimation, and a real-time algorithms using Kalman filter isproposed. The simulation results prove the validation. Thus, problems based ontwo camera visual system have been studied, which proves the way to theresearches of advanced robot.