The Research On Binocular Parallax Model And Its Application In Visual Servoing

Visual servoing is a technique which uses feedback information extracted from a visual sensor to control the motion of a robot. Visual Servoing techniques are basically classified into two types, position based visual servoing (PBVS) and image based visual servoing (IBVS). PBVS is sensitive to the calibration error on camera parameters and robot structural parameters. IBVS is based on the error between current and desired features in the image space, and does not involve any estimate of the pose of the target. But the image based visual servoing involves estimate of the Jacobian matrix, which describes how image feature parameters change with respect of changing manipulator pose. The chief advantage of IBVS over PBVS is that the positioning accuracy of the system is less sensitive to camera calibration errors. It is difficult to estimate the Jacobian by using classic binocular vision model. To simplify the Jacobian estimating or servo control designing, this work presents a binocular parallax model with the configuration that the two cameras is symmetrical and the optical arises intersected. In this work, the binocular parallax model is employed to design the servo controllers for two applications, DIM and iTracer. The remainder of this dissertation offers the following:1. A symmetrically configured binocular parallax model using geometric relations in space, which maintains a proper field of view and eliminates the need for data training, is proposed. As it allows for decoupling and linearizing, this proposed binocular parallax model greatly simplifies visual servo implementation, has small settling time, and is insensitive to modeling errors.2. A very efficient algorithm was proposed for fast template matching. The coarse-to-fine strategy was adopted in this algorithm. At coarse searching stage, Local entropy difference was employed for template and sub-window of characterizing. Moreover, integral image and square integral image of the input image were utilized to speed up the local entropy difference calculating. After obtained the local entropy differences for every position within input image, the candidates table was built by threshold determining. At the fine searching stage, a new matching criterion, called as combine correlation, was introduced for the best similar position within candidates table.3. A DIM is designed for the instruments pointing and positioning in ICF experiment. The DIM is formed from a hybrid manipulator and provides precision2-rotation and1-translation capability for positioning of diagnostic instruments. The inverse kinematics and close-form direct kinematics of the DIM are presented. Moreover, the error mapping function, which represents the end positioning error resulted from limbs length error and joint clearances, is formulated by using matrix method with the differential of kinematics function. Moreover a3DOF robotic platform is designed. Together with the binocular vision system a3DOF eye in hand system is implemented. Furthermore, the kinematic and dynamic model of the robotic manipulator is derived, and a PID controller is designed for target tracking in3D space.4. Employing the binocular parallax model to design the servo controllers for DIM and iTracer. The DIM guides the Instruments locating at a desired position with the parallax vector as feedback. And iTracer track and lock a moving target in task space with the help of parallax state and kalman filter.