Study On Modeling And Calibration Of Portable Vision Metrology

In this paper, a monocular vision inspection system was designed for thedetection of rotator profile. To achieve the imaging of large-scale geometriccharacteristics, a mobile control mechanism was brought in to drive the imageacquisition system. According to the special structure of the inspection system, thispaper established its mathematical model and analyzed its model error. A method wasproposed for the realization of accurate measurement: firstly, calibrate the camera’snonlinear parameters; secondly, correct the distorted image using the calibratednonlinear parameters; finally, solve the information of untested geometriccharacteristics from the corrected image.According to the system development procedure, the design of system structure,the establishment of the mathematical models and the calibration of systemparameters were presented in turn. The main work of this thesis is as following:1) According to the analysis on the rotator structure and detection accuracy, theinspection system is divided into two parts: the image acquisition system and mobilecontrol mechanism. When building the image acquisition system, we selectedbacklight illumination, telecentric lens and high resolution camera; the imaging oflarge-scale geometric characteristics can be completed with the auxiliary of mobilecontrol mechanism. Experiment show that clear images can be photographed at everyrotator position with the precisely drive of mobile motion mechanism.2) Based on the introduction of camera linear model and nonlinear model, thecamera’s model of this article was proposed and the imaging error caused by weakperspective model was analyzed. Calculation results show that the imaging error isless than0.0055mm, which can meet the need of the detection accuracy.3) To achieve the precision measurement of large-scale geometric characteristics,the calibration of inspection system was divided into three steps. Firstly, solve thecamera nonlinear parameters according to the relative position relation of the sametarget points in two pictures; secondly, revert the ideal image from the distorted imageusing the nonlinear parameters calibrated; finally, solve the camera linear parametersaccording to the coordinate relation of target points with their image points.Simulation experiment shows that average error of distortion center is (0.2243,0.1636) pixel, and distortion coefficient error is0.28%at the circumstance that target pointsnumber is25, and noise level is0.2pixel. Real image experiments confirmed theeffectiveness of nonlinear parameters calibration method and the stability of linearparameters calibration method. The measurement principle of large-scale geometriccharacteristics was presented after the calibration of system parameters.