High resolution vision-based servomechanism using a dynamic target with application to CNC machines.

摘要

This dissertation introduces a novel three dimensional vision-based servomechanism with application to real time position control for manufacturing equipment, such as Computer Numerical Control (CNC) machine tools. The proposed system directly observes the multi-dimensional position of a point on the moving tool relative to a fixed ground, thus bypassing the inaccurate kinematic model normally used to convert axis sensor-readings into an estimate of the tool position.;A charge-coupled device (CCD camera) is used as the position transducer, which directly measures the current position error of the tool referenced to an absolute coordinate system. Due to the direct-sensing nature of the transducer no geometric error compensation is required. Two new signal processing algorithms, based on a recursive Newton-Raphson optimization routine, are developed to process the input data collected through digital imaging. The algorithms allow simultaneous high-precision position and orientation estimation from single readings. The desired displacement command of the tool in a planar environment is emulated, in one end of the kinematic chain, by an active element or active target pattern on a liquid-crystal display (LCD). On the other end of the kinematic chain the digital camera observes the active target and provides visual feedback information utilized for position control of the tool. Implementation is carried out on an XYthetaZ stage, which is position with high resolution.;The introduction of the camera into the control loop yields a visual servo architecture; the dynamic problems and stability assessment of which are analyzed in depth for the case study of the single CAM- single image processing thread-configuration. Finally, two new command generation protocols are explained for full implementation of the proposed structure in real-time control applications. Command issuing resolutions do not depend upon the size of the smallest element of the grid/display being imaged, but can instead be determined in accordance with the sensor's resolution.