Abstract

A vision-based pipeline robot is studied in this paper, and a strategy of vision navigation is proposed. The key technologies such as anomaly determination of pipeline inner wall, curve identifying and multi-sensor fusion are investigated. The movement of pipeline robot is controlled by robot vision, at the same time Kalman filters are expanded to fuse measurements of multi-sensors to identify the robot in the environment of straight pipe, elbow pipe and T-tube, so as to realize the navigation and positioning of the robot in the pipe. The TCP/IP is used to provide the Socket programming interface to realize remote information exchange for the pipeline robot, and realize the data communication between the host computer and the router. According to standard experimental detecting on pipeline section, the method can meet the detecting requirements.