Research And Application Of Binocular Stereo Vision For Blank Size Measurement

In the field of mechanical manufacturing automation,the use of advanced three-dimen sional measurement technology to quickly and accurately detect the size of the blank has important guiding significance for the accurate determination of the cutting volume in the reprocessing process.At present,most traditional measurement technologies cannot meet the measurement requirements of modern manufacturing enterprises.Vision measurement technology has been widely used due to its non-contact,high degree of measurement accuracy,and fast speed,and has become a three-dimensional detection technology with broad prospects.Based on the principle of binocular stereo vision measurement,this paper studied the key issues of image acquisition,camera calibration,image segmentation,and image matching for metal blanks,and designed a software and hardware system that can obtain the 3D information of the feature points of the blank and achieve the corresponding 3D dimension.First,The paper selected the hardware devices required for the binocular vision meas urement system,mainly including industrial cameras,lenses,and light sources;completed the design of the hardware part of the image acquisition system;and then aiming at the problem of low calibration accuracy in camera single target calibration,a new method based on pinhole imaging model and immune particle swarm optimization algorithm was proposed to optimize the initial calibration parameters of the camera.Experiments showed that the calibration m ethod has good feasibility and accuracy,and can obtain higher calibration accuracy,Furthermore,more accurate dual-targeting results could be also obtained.In addition,for the sake of reducing the mismatch rate of the later image matching,the epipolar constraint method was used to correct the camera parameters.Aiming at the highly reflective characteristics of the metal blank,an image segmentation algorithm based on the division operation and color model conversion combined with the immune algorithm was proposed.This algorithm could well extract target workpieces from complex backgrounds.,experiments proved that the algorithm proposed in this chapter had strong denoising ability,rapid speed and high accuracy,and could improve the accuracy and speed of online real-time detection of workpieces.By analyzing the existing stereo matching methods,the rough matching part and the fine matching part of the feature matching method were studied respectively.In image matching,the classic matching algorithm was improved in rough matching and fine matching,respec tively.The proposed stereo matching algorithm could obtain more correct matching point pairs,thereby ensuring a high acquisition amount of three-dimensional size information of the rough parts.Finally,the least square method was used to obtain the three-dimensional coordinates of the feature points on the blank,and then the corresponding part size and depth information was received.Based on the above research,corresponding algorithm programs such as camera ca libration,image segmentation,and image matching were written,and debugging was perf ormed using Matlab2014 a software to verify the feasibility and effectiveness of the algorithm used.Finally,two actual shot blanks were selected for measurement experiments.It could be seen from the final experimental results that the binocular vision measurement system proposed in this paper could effectively obtain the three-dimensional coordinate values of the feature points on the blank and the corresponding three-dimensional dimensions.