Research On The Technology And Application Of Stereo Vision For Mobile Robot

Mobile robot is an intelligent robot, which can prceive the environment and its own states byexternal sensors, so it can analysis the environment and make decisions independently,andachieve specified functions. The most important technology of mobile robot in motion isself-navigation and positioning. Among all the navigation technologies, using stereo vision formobile robot navigation, obstacle avoidance, is one of the main directions of development ofmobile robots navigation.Arming at the demanding for real time and accuracy of stereo matching in vision-basednavigation,we choose dynamic programming as our stereo matching algorithm. But theperformance of conventional DP has not been satisfactory and horizontal streaks occur whenapplied to the stereo matching problems.To solve these problems, in this paper,first we propose anenergy aggregation method which can use the information of object boundary to determine thedisparity changing regions. So we obtain a highly accurate initial disparity map which can makethe subsequent phases of parallax getting get a good performance.And we propose a new stereomatching algorithm using the constraints in horizontal, vertical directions and in the direction ofthe edges of objects as the smoothness. For the the boundary pixels of the image,Constraining thecurrent pixels by use of previous matching information to strengthen the object boundary pixelsdisparity discontinuities.We apply the boundary contour information and multi-directionalconstraints to the choice of state transitions of tri-state DP.And we also introduce the idea ofmulti-path backtracking to exploit the information gained from DP more effectively. Theexperiment shows that this method alleviate the effect of horizontal steaks, mismatches in thedirection of boundary and has a high matching accuracy and a fast speed.At last, we apply the stereo matching algorithm we proposed into vision-based navigationsystem based on AS-R robot platform, we get the photos of the scenes from a binccular camera,and then realize the function of obstacle avoidance and navigation. Experimental results show thatthe application of the algorithm for mobile robots in indoor environments are effective andfeasible,and we achieved the stated objectives.