Study On Calibration Method Of Micro Stereovision System And Half-occlusion Of Microstructure

Micro stereovision system based on stereo light microscope (SLM) has been applied insome micro-domains widely. Highly accurate measurement and location are implemented bythe vision feedback. Sequentially, the micro-actor is controlled to complete the expectedoperation. Half-cclusion is an important influencing factor in stereo matching and 3Dreconstruction. To analyse half-occlusion of microstrucmre in SLM, vision system iscalibrated firstly, whose intrinsic and extrinsic parameters can be acquired subsequently, thenanalytical models are built, finally, half-occlusion of mircostructures including microchannelare analysed, in mircro stereovision.The purpose of calibration of micro stereovision system is to get the intrinsic andextrinsic parameters of the corresponding vision model. For reference of Zhang ZY's linearvision model calibration theory and consideration of micro stereovision character, a two-stepcalibration method is put forward. The study on calibration is as follows:①Data source arecomers which are extracted from many images using Harris arithmetic operators. First, only aproportional value namely magnification is acquired from elementary calibration since SLM'sdepth of focus is finite and the calibration pattern can only rotate in a plane, then, forthe constraint that operating distance is constant, foci can be got, then intrinsic parameters canbe achieved; second, extrinsic parameters can be acquired from the further calibration;②Based on SLM micro stereovision system, a set of calibration device system was assembledand calibration patternⅡwas designed. Some components were levelling and collimated. Thefollowing experiments verified the practicability of this system;③The average, standarddeviation and the relative error of foci, ubiety and magnification of binocular system aregained by the calibration of the CMO-SLM under three common-used objective lensmagnifications (2.0X, 2.5X, 3.2X). The relative errors of magnification and foci are less than1.7 percentages, while the relative errors of the fitting straight lines' rate of grade which isclose to 5 and a criteria for evaluation of the angle around z axis are less than 0.6 percentages.The relative deviation between left and right translation vectors can judge the accuracy of thetranslation vector, which is less than 0.116mm.Half-occlusion of the microstrucmre in the G-SLM and CMO-SLM are formulatedseparately, after that experimental scheme is presented and primary experimental verificationis completed. The study on occlusion is as follows:①Analytical vision models ofhalf-occlusion are built. G-SLM analytical model is derived from the coaxial spherical planes lens model while CMO-SLM analytical model is from the common-used pinhole model;②Field of vision and the relational expressions between coodianates of object point and criticalocclusion angle which depends on included angle, foci, effective width of CCD areformulated;:③Occlusion of microchannel is analysed weightily. As to ideal and nonidealmicrochannel which have fixed rake angle of inwall, half-occlusions are explained byanalytical vision models, after that quantitative conclusions are achieved;④Study onexperimental method which can verificate relational expressions in③is completed, after thatprimary experiments are done. Microchannel's continuous rake angles are simulated bycalibration pattern and a microchannel which has no occlusion and was hot embossed withsilicon mould. The average of absolute error is 1.04°;⑤The range of depth-to-width ratio ofmicrochannels which have no opposite side occlusion after some formulation.Calibration is a prerequisite of analysis of half-occlusion, while the correspondinganalysis lends itself to half-occlusion solution.