Research On Computer Vision Based Mobile Robot Control System

Computer vision based mobile robot navigation is one of the hot topics in robot field recently. Just like humans, vision is also the most important source for robot to know the world around; it can supply the information of the environment, such as the objects'distance and direction. It is significant to research on mobile robot vision system and the related technology, such as camera calibration, multi-sensor fusion, robot navigation and control theory. Color vision signal is the basic information for mobile robot to "feel"the world around it. So, it is the key for mobile robot to get the information of the environment from the vision signal rapidly and correctly; the vision system is the basis of the control system. This dissertation is devoted to the research of the digital image process, the design and realization of the mobile robot vision control system. First of all, the color model is discussed to present the color recognition theory for mobile robot. Some popular color space models (RGB, YUV and HSV) and the converting formula between them are introduced. The relation and difference between them are compared in both picture and data sheet. The image segmentation and clustering algorithm are discussed and used to detect objects in images. A method of using double threshold of H value to segment the object from the background is used in this essay. A fast scan line seed fill based clustering algorithm for two-value image is proposed to get the objects'rectangle areas. To get 3-D information from the 2-D image is the ultimate goal for computer vision, and it is also the crucial task for robot to locate itself. The camera projection model and distortion model are presented; several well-known calibration methods are introduced in this essay. A self-calibration method is presented to get the interior and exterior parameter of the camera. In the lab's condition, the distance and the direction of the object can be computed on the model of perspective projection. Single-camera and local vision based mobile robot location and navigation system is achieved in this essay. Finally, the framework of the whole control system is discussed. A LAN based control structure is adopted. The image processing and control algorithm are fulfilled on the platform of PC; the control command are sent to the server in the robot through socket programming; the server program is in charge of the status collection and command execution. DirectShow technology is used to capture and process the video signal, and ARIA developing library is used to control the robot.