Monocular Vision Navigation System Design Of Robot Based On Ground Trajectory

Trajectory detecting and tracking on the ground is an elementary technique in robot-vision guided systems. In this thesis, a monocular vision navigation system composed of camera, Digital Signal Controller (DSC) and mobile agency driven by stepper motors is built, which can realize real-time tracking according to JPEG images in the case of unknown rail line parameters.For JPEG image decoding, a fast two dimension 8×8 Inverse Discrete Cosine Transform (IDCT) algorithm based on Look-Up Table (LUT) is used. This algorithm avoids multiplications by using LUT and decreases additions greatly by utilizing the characteristics of actual images and the symmetries of the basic images. And using an improved median filtering algorithm to eliminate the image noise, which can effectively reduce the time complexity and meet the requirement of real-time image processing without sorting.In order to effectively solve the real-time problem of extracting reference trajectory, according to the feature that the width of the trajectory keeps almost the same in the image, a fast line extraction algorithm is presented. Combining Elementary Line Segment (ELS) method to improve Hough transform, and then obtains the mapping parameters in parameter space of the trajectory equation.Input of the state-based tracking control is achieved by extracting trajectory direction deviation parameters and distance deviation parameters from images captured by camera. According to the value of the direction deviation parameters, two different control strategies are adopted. In order to improve the control performance, when the image distortion is too large resulting from camera obliquity, these parameters are transformed from image coordinate to robot coordinate through projective transformation, while parameters such as robot position and trajectory curvature are not necessarily provided. Experimental results demonstrate the validity of the system that can realize real-time tracking to arbitrary shape trajectories, which has achieved the expected results.