The Research Of Mobile Robot Road Environment Analysis Based On Visual Information

The outdoor mobile robot’s work environment is complex, and has varied tasks to complete, which has proposed higher requirements for the robot accuracy: more accurate positioning, more suitable speed, more comprehensive perception, more efficient operation. The conventional sensors(a distance sensor, a sound sensor, etc.) has been unable to fully feedback to the robot the environmental information, such as prediction of what type of material the road will go through, determination of the black areas on the road surface is fallen leaves, water or shadows.More than 70% of the human’s experience information of objective world comes from eyes, in terms of robot, the visual system is the robot’s eyes. It detects the amount of informations which response the most comprehensive and in-depth environmental. So it’s considered the most important perception force. Therefore, for outdoor mobile robot, the visual navigation is one of the most critical science and technology.The purpose of the research is to collect the environmental image information and extract and analysis the road features,then identify the type of road environment and provide the informations to the robot as the supplementary to it’s path planning during moving in the complex environment of unstructured road. According to study other researchers’ result of four common environmental recognition experiment, the final selection of the feature extraction from the two aspects of environmental topography color and texture. Then compare the various extraction methods and experimental results of the four kinds of texture features and five kinds of color features extraction algorithm and then finally decided to adopt the method of the first-order center distance of HSV color space to extract color feature and the method of GLCM for texture feature extraction. According to the two methods of multiple images of three typical experimental environment topography of color and texture feature extraction and establish characteristic sample database.Finally, by the use of Matlab programming to train the experimental sample library and do the offline and online classification test on the environmental pavement. As the experimental results shows, the method this subject proposed has a high accuracy rate, which verifies the road identification method proposed feasibility and reliability of this project.