Research On Moving Object Tracking Based On Binocular Wheel Mobile Robot

In recent years, with the continuous development of computer vision technology andmobile robotics technology, researchers show a greater expectation that the robot’s eyes canbe compared with human beings’ in achieving the perception and understanding the scene.Learning more about computer vision, we can introduce the three-dimensional information ofthe scene into a two-dimensional image and the moving target tracking can be more widelyused in robot navigation and positioning, intelligent human-computer interaction, virtualreality, and imaging guidance.In this paper, a combination of stereo vision technology and tracking technology is usedin the three-dimensional real-time tracking of moving targets based on the binocular mobilerobot platform. Some research are carried out in Binocular camera calibration, stereomatching for disparity graph, and also moving object extraction and tracking.In this paper, the dual-plane camera calibration method based on the introduction oftangential distortion establishes a more comprehensive model of camera calibration. Putforward a nonlinear optimized algorithm for solving stereo parameters which can reduce theaccumulation of external parameter errors on the calculation accuracy. Study a simple andeffective epipolar rectification method in the calibrated case, which lays the foundation toimprove the precision of stereo matching. The fast matching method based on gray similarityintroduced the box filtering optimizes the calculation of SAD matching cost; using epipolarconstraint, uniqueness constraint and continuity constraint to narrow the scope of thematching search; using error evaluation function and BM fast matching to select the bestmatch point. Finally, the comparative experiments with the SMP matching algorithm provethe advantages of this method.This paper presents an adaptive color image segmentation method to extract the targetarea quickly and accurately. Studied the linear and nonlinear calculation of thethree-dimensional point, linear algorithm of distance measurement combined with the targetparticle extraction realizes the three-dimensional measurements of the target, and experimentsshow that the method has strong robustness and high computational efficiency. A3D trackingmethod based on the binocular vision mobile robot was proposed, which uses Kalman filter asmotion estimator. The method could predict the target position accurately because ofintroducing both motion information of the robot and three-dimensional information of thetarget as a control inputs to achieve. The target position as the rotation and straight of the robot movement is better in tracking moving targets and improving accuracy of the distancemeasurement based on binocular vision.The last part is the analysis of hardware systems of the binocular wheel mobile robot,and designs a real-time software system of3D target tracking with using standardprogramming languages and modular design, which has strong cross-platform portability.