Grid Map Building Based On The Sequences Of Omni-Directional Images

Based on the project of lunar rover, a vision system for the automatic guidance of an explorer rover is developed. The vision system is aimed to reconstruct a two dimensional grid map using sequences of panoramic images obtained by an Omni-directional camera, which is used to realize the function of path planning. First, a calibration process is needed to calibrate the distorted panoramic images. The sequence of distorted images will be turned into bird’s eye view images after calibration. The second step of the image process is to segment bird’s eye view image into different areas. According to the segmented areas’color character, the areas are classified as passable and impassable areas. Then a local grid map can be reconstructed based on areas’character. And artificial landmarks for localization can be recognized by color. Each three different landmarks make up a landmark group whose features are used to do landmark matching in two images. The localization can be realized with three landmarks. Finally, the sequential of local grid maps are put together and overlapped according to the explorer’s position to obtain the global grid map.The hardware of this Omni-directional vision system is made up of a Sony Omni-directional camera and an image grabber card. The software is developed under VC6.0 environment. The vision system cooperates with the dual-camera vision system which is used to avoid obstacles in time, realizing the automatic guidance of the explorer.