Design Of Digital Binocular Stereo Microscope System

The development of industrial automation promotes the continuous expansion of the measurement field to high-efficiency,precision,and microcosmic aspects.In recent years,microscopic stereo vision has been widely used in microscopic three-dimensional measurement,and has good detection results in industrial inspection,medical,microelectromechanical and other fields.Microscopic stereo vision is a technology that uses stereo microscopes and two cameras to obtain binocular images from a microscopic perspective and then perform stereo matching to achieve 3D reconstruction of the target.This paper takes the microscopic stereo vision system as the research object,and studies a digital stereoscopic microscope system with simple structure,low cost and high measurement accuracy and measurement speed.First,build a basic microscopy system using common stereo microscopes and two cameras on the market.According to the size of the microscope eyepiece and camera,design a detachable eyepiece and camera connection to achieve a coaxial,adjustable distance connection.After the design of the mechanism position adjustment,the study of the relationship between image change and displacement,the calibration of nonlinear parameters and the adjustment of image parameters,appropriate binocular images are collected.Secondly,research on efficient and accurate stereo matching algorithm based on the CUDA architecture platform.Including the steps of binocular image alignment and preprocessing.In addition,two improvements are made to the commonly used SGBM stereo matching algorithm.One is to propose a new composite BT cost to improve the edge detail extraction of multi-channel images and enhance the matching features,the other is to propose based on The semi-global matching algorithm of the image pyramid performs cost aggregation to obtain the disparity.And at the same time improves the efficiency of binocular image matching,and then performs post-processing through median filtering and confidence detection to obtain a good disparity map.Then convert the obtained disparity map into a depth map,and use the chip and capacitor as the target object to quantitatively analyze the relationship between magnification-disparity-accuracy,and analyze the limitations of the system through image processing in different lighting environments.At the same time,the high magnification objective lens and eyepiece were replaced to study the error range and error reason under 7.5X to 180 X magnification,and the robustness and accuracy of the system and algorithm were determined.The comparison with other research results also shows that the system has higher advantages in accuracy and speed.The studied binocular stereo microscopy system can achieve microscopic measurement with a maximum error range of 2μm.For a 1024*768 resolution binocular image,three-dimensional measurement data can be obtained within 0.1s.The point cloud image generated from the depth map can be visually observed three-dimensional reconstruction effect.This system can quickly realize application development based on in-depth inspection.The paper takes the in-place detection of the wiring harness terminal as an example,combined with the human-computer interaction interface written by CUDA and Qt,extracts the depth of the corresponding detection area through the image morphology method,and quickly realizes the terminal detection application.