Research On Obstacle Detection Of Mobile Robots Based On Multi-sensor Information Fusion

The detection and localization of obstacle is the key and hot issue in the research field of the mobile robots, and it also becomes the principal issue of avoiding the obstacles and map building.This dissertation aims at designing an obstacle detection method based on the information fusion of laser radar and camera and takes the mobile robot MORCS-1 as experiment platform which has been developed by the Intelligent Laboratory of Central South University. The main contribution and work are described as follows:1. It puts forward a new monocular obstacle detection method based on straight lines matching. First we extract edge points with Sobel algorithm in a single image, and we assemble edge points to straight lines according to some rules, then we match the lines using the properties of themselves and gray value of pixel in the image, and reconstruct obstacle contour according matching result. Finally we locate the dubious obstacles with the theory of monocular measurement of distance.2. It puts forward a new method of obstacle detection based on information fusion of laser radar and camera. Firstly, we figure out the horizontal angle and vertical angle with which the laser radar can detect the dubious obstacle in terms of the monocular detection. Then we can judge whether the dubious obstacle is real obstacle in light of the information of laser radar retrieving with the horizontal angle and vertical angle, and we can accurately locate the obstacles.The experiment proves that the method proposed in this essay can detect obstacle more accurate and reliable than single sensor, and it can meet the requirement of real time.