Research On Networking Strategy And Optimization Method Of Industrial Stereo-vision Measurement Network

Stereo-vision measurement is widely used and researched in the industry,including product inspection,assembly and reverse engineering and so on.Stereo-vision network measurement is a technical guarantee for the realization of 3D measurement,which determines the speed of measurement and the accuracy of measurement results,and has great engineering significance in practical applications.This paper did a deeper study of stereo-vision measurement networks.Firstly,Based on the study of the mathematical model of industrial stereo-vision measurement theory,this paper proposes a optimal method of stereo-vision measurement network.A model of the object is obtained by the Structure From Motion algorithm.According to the model of the object,the reference definition is used to study the camera position constraints and estimate the optimal distance between the camera and the object.Combining the geometric properties of the ellipsoid model,a large number of initial viewpoint positions around the object to be tested are designed.Then based on industrial multiple-viewpoints stereo-vision 3D reconstruction method and visibility constraints,a screening method based on cluster analysis and loop iteration is proposed to screen the initial viewpoint position.The least number of viewpoint and satisfactory images are achieved.The image point error physical model,cameras' external reference calibration theory,and cameras' internal and external parameter calibration methods are studied.The camera at the viewpoint position in the obtained stereo vision measurement network is calibrated.Finally,the lampshade is used as the measurement object.When the F numbers of the cameras are f/1.4,f/2,f/2.8,f/4,f/11,f/16,respectively,the ellipsoidal imaging network and the spherical imaging network are compared and analyzed.The lampshade is performed 3D reconstruction.The result shows that afterscreening,on the basis of ensuring coverage and measurement accuracy,the number of network nodes is significantly reduced.In other words,the number of camera images is significantly reduced.It also shows that the ellipsoidal imaging network is superior to the spherical imaging network,which verifies the feasibility of this method.This paper uses machine vision to achieve three dimensional measurement,which improves measurement speed while ensuring measurement accuracy and coverage,and provides theoretical and technical support for industrial measurement.