Design And Implementation Of Mobile Robot Simultaneous Localization And Mapping System Based On Binocular Vision

The problem of SLAM(Simultaneous Localization and Mapping) is a basic problem in research field of mobile robotics as well as a key step to implement the independent navigation and autonomous control system. With the development of image processing and computer vision, visual sensors has been widely applied in research of SLAM. It is replacing other sensors and becoming the main device to collect the information of environment. It has many advantages, such as low cost, less energy-consumption, mass information and easy operation. In this paper, how to implement the SLAM system by binocular vision is studied and how to design and implement the SLAM system is described. The main research contents include:First, this paper studied the mobile robot SLAM system, including its history, development, significance of research and explained the mainly problems which have to be solved during research. Problems which will happend during implementation are stated. Then, the motive model, pose model and observation model, which are the basic parts in SLAM system, are built and analyzed.Second, in order to implement the range-measurement system by binocular vision, this paper analyzed the basic theories, such as epipolar geometry, epipolar constraint and theory of triangulation measuring. The main steps to implement range-measurement system include camera calibration, stereo matching, feature points detection and matching and triangulation measuring. In the experiment, the algorithm of SIFT(Scale-invariant feature transform) was used to detect the feature points and epipolar constraint was used to reduce the probability of mismatching. So, it is reliable and high-efficiency when detecting the feature points by SIFT and epipolar constraint. In order to improve the efficiency of detecting and matching, this paper put forward two solutions, one is extract the region of interest and the other is improve the SIFT, then both of them have been proved that they are effective in experiment.Third, aiming at how to solve problem of estimating states of robot’s pose and environment’s features, this paper analyzed the advantages and disadvantages of some kinds of algorithm. Through these analyses above and the simulation results of Fast SLAM 1.0 and Fast SLAM 2.0, then, Fast SLAM 2.0 was found that it has more advantages than other algorithm for the SLAM system. So, Fast SLAM 2.0 was chose as the algorithm to estimate state parameters.Finally, mobile robot SLAM system with abilities of moving, detecting pose information of itself and collecting environment information was designed and implemented. A small two-wheel-drive vehicle was chose as a main part of the mobile robot. The range-measurement system based on binocular vision and control system of the robot was designed and finished. Then, environment information was collected and transmitted into the upper computer to be processed, the SLAM system was implemented.