Image Information Detection and Integration of Large Load Visual System
|Course||Detection Technology and Automation|
|Keywords||Camera calibration Genetic algorithm Motion estimation Optical flow estimation Vision sensor Binocular matching|
In the large load system, the control of little displacement with high precision is an unresolved problem. In this thesis, the control platform of the large load system is built based on machine vision theory, the sampling is actualized by images at different moments on the one hand, the monocular stereo vision sensor system is designed with four lenses to extract the motion information on the other hand. The motion information of the large load system is further analyzed to realize its control.Camera calibration is necessary to construct the visual platform and is also essential to get information from 2D to 3D. Image pretreatment is a preparative to analyze images. In computer vision, two distinct approaches of extracting motion information from images have been developed: the feature based and optical flow based approach. Both of them correspond respectively to the short time with few displacements and the long time with big displacement. The optical flow algorithm is applied to the large load system and the information of 2D and 3D is then obtained. At last, the monocular stereo vision sensor is designed base on the character of the large load system and its mathematical model is established and the geometrical parameters of its structure are analyzed. It offers the theoretic tool to optimize the stereo vision sensor.The essential works of the thesis can be summarized as follows:1. The approach of Camera calibration is introduced that the linear equations join the nonlinear aberrance equations. The camera parameters are optimized by combining least square algorithm and the GA.2. The Wiener filtering is improved by integrating the nonlinear ordination. Then gradient operator is used for edge detection of images.3. The thesis adopts optical flow method to estimate 2D and 3D motion, and implements 2D motion estimation by gradient operator. The parameters of 3D motion and the structure are sequentially estimated.4. The stereo vision sensor system is designed for large load system, its mathematical model is established, and the geometrical parameter of the structure is then analyzed. The geometric parameters of the stereo vision sensor system are finally confirmed.