UBC Theses and Dissertations
Torso-mounted visual servoing of an outdoor robotic manipulator Ziraknejad, Nima
In the majority of vision applications, sensor calibration is a prerequisite to proper use of the sensor for both measurement and control. The objective in camera calibration is to estimate a set of parameters to construct a mapping between the 3D position of a target point and its 2D image coordinates. This thesis presents an autonomous stereo camera calibration technique based on a pinhole camera model with applications in industrial outdoor visual servoing systems. An efficient iterative least-squares parameter estimation algorithm is used to estimate the camera model parameters. The obtained stereo camera model is used to estimate the pose of the target object during the robot servoing process. The heavy-duty stereo camera rig is installed on the torso of an outdoor 3DOF robotic manipulator. The stereo camera calibration is entirely an autonomous process as the robot moves the calibration tool within its workspace and the stereo camera model is produced after the data collection process. The stereo cameras are treated as a single unit and a single transformation is obtained for the stereo camera pair in the system. The calibration process is fast, efficient and no human interaction is required during the process. The developed vision system is capable of detecting the 3D positions of the objects within the robot workspace with an average error of 1.93cm (maximum error of 7.6cm) for a region equal and smaller than the region in which the camera calibration is performed. The errors are comprised of errors in camera calibration, robot positioning and image processing. The maximum error represents an error of 2.1% of the total robot workspace reach. The visual servoing tasks are being performed with an update rate of 29ms. The complete visual servo system, used for pipe handling, has been implemented on a combined platform with two main processing units and a fully integrated 3DOF torso-mounted robotic manipulator. The processing units include an industrial PC104 embedded system with 900MHz Pentium III CPU for vision related tasks and an industrial Programmable Logic Control (PLC) unit for low level control of the robot.
Item Citations and Data