Vision-Based Self-Propelled Vehicle Road Recobnition And Autonomous Navigation

Mobile robot research has broad application prospects in space exploration, military, civil and other areas. It also provides a good development platform for a variety of disciplines, such as artificial intelligence, data fusion and information processing. Computer vision-based autonomous navigation technology is an important direction in intelligent robotics research fields.In this thesis, we research a set of self-designed vision-based self-propelled vehicle system. In the structured road environment, the self-propelled vehicle uses a CCD camera as a vision sensor. It will collect the road information through the vision sensor. Then, through a PC, real-time image processing will be done to identify the road and make judgments to control the self-propelled vehicle straight ahead along the lane line or taking turns. This thesis discusses and compares some related image processing algorithms. After doing contradistinction and analysis to the experimental results, it proposes the two-dimensional linear median filtering method and the Sobel edge detection method for this thesis' research. In the lane line identification, the thesis uses the known point Hough transform to extract a straight line, uses sub-identification method to detect bend lines and uses least square method to fit lane lines. Self-propelled vehicle autonomous navigation can be divided into straight-line tracking and turn lane turning control. Bias control has also been discussed in this thesis. In the end, self-propelled vehicle system has good performance in the road detection results and algorithms efficiency, reaches the autonomous navigation experiment results.