ALV Research And Implementation Of Segmentation Algorithm Of Two - Dimensional Road

Autonomous Land Vehicle (ALV) is an intelligent mobile robot, which can run on the road or on the off-road scenery autonomously and continuously. Its investigation, which involves theories and technologies of many disciplines, and embodies the latest achievements in information science and artificial intelligence, is of great value in research and application. Therefore, it is paid highly attentions by all over the world. Of all the ALV key techniques, vision-based navigation perceives and understands all the surroundings where ALV runs quickly and accurately, and then determines a safe path for ALV. The key to the vision-based navigation is the rapid road detection. Therefore vision-based navigation is one of the key investigations on ALV.In this paper, we mainly researched on the ALV 2D vision-based navigation, and furthermore realized the 2D road image-perception algorithm. To detect road area accurately, robustly and timely, based on the other people's research, we proposed an image segmentation algorithm, which based on the color histogram and clustering technology. Additionally, we made some relative optimization on the algorithm's complexity in space and spectrum. It is proved by experiments that this algorithm can achieve good result in common road environments. When the algorithm is running on the server with double Xeon 2.4GHz CPUs and lGbytes RAM, its processing time is less than 40 ms per frame.Shadow is an inevitable interference to road detection. The effect of shadow to image segmentation is notable. If processed improperly, shadow will make negative effect to image segmentation and consequently, the navigation. Therefore, according to the characteristics of shadow on the road, we proposed a novel method to detect and remove shadow. This method makes use of not only image segmentation but also edge information, which is available through edge detection and Hough transform. It contains the two kinds of information mentioned above to get the final road edge. It is proved that this algorithm works well enough to the road withshadow, and it also achieve the demand of process time.