A new algorithm is proposed for fast generation of the environment map from stereo images as imaged by two cameras on an indoor mobile robot. In the new algorithm, the environment map is made from reduced images.
The amount of computation can be dramatically reduced if reduced images are used instead of original images. Image reduction is the process which reduces the size of images by averaging a given region on the original images. The more images are reduced, the worse accuracy is obtained in a map. To keep the accuracy of original images, the original images were referenced during line fitting process. To know the accurate heading angle of mobile robot is essential to detect significant lines precisely. A reliable calibration algorithm for heading angle is presented, which uses both vanishing points and odometer information. And, to reduce the overall processing time, the presented algorithm is reformed into distributed processing version.
The new algorithm is compared with the algorithm which uses original images. The experimental results show the proposed algorithm saves processing time, generates a reasonably accurate map and satisfies the requirement of realtime processing for the given specification.