Research On Precision Control Technology Based On Binocular Stereo Vision System

Binocular vision measurement system has become one of the most convenient and effective measurement methods in the three-dimensional measurement of large-scale equipment assembly or structural parts in aerospace,military and other fields.But the main problems of current vision measurement are: its measurement accuracy is limited and it is difficult to improve,which has become one of the main bottlenecks restricting the development of vision measurement technology.There are many error factors of measurement accuracy and non-linearity measurement model,which lead to the complexity of conventional error analysis and the difficulty of vision error analysis.Aiming at these problems,this thesis proposed a feasible error analysis method based on binocular vision measurement system,and the main work are as follows:(1)The structural composition of the binocular vision measurement system is analyzed.From the establishment of the binocular stereo vision structure and the three-dimensional reconstruction mathematical model,the image imaging point coordinates,the effective focal length of the camera and the conversion relationship between the camera coordinate system are analyzed and analyzed.The measurement method of the spatial target feature point in the binocular vision measurement coordinate system.(2)Aiming at the problem of many error factors and complexity of error analysis model in binocular vision system,the error factors in binocular vision system are summarized based on binocular vision measurement structure,system layout and measurement environment.The main error factors such as calibration error,feature point extraction and matching error are also analyzed.The experimental results show that the method of parameter calibration with high evaluation index can effectively restrain the influence of various error factors on the final measurement accuracy.It also proves that independent distortion correction can improve the calibration accuracy of binocular vision measurement parameters.(3)Aiming at the problem that lens distortion seriously affects the accuracy of other parameters,a point-by-point distortion correction method based on virtual template is proposed.This method separates distortion correction and parameter calibration,which are closely related to the measurement accuracy of binocular vision system.Firstly,the distortion is corrected independently,and then the camera's internal and external parameters are calibrated on the basis of undistorted images.This method does not need stereo distortion correction template with high processing accuracy,precise measurement distance between camera and correction template,and the camera's internal and external parameters calculated by vision imaging model.The final distortion correction data table can fit distortion correction for any calibration or measurement image without parameter calibration or visual measurement at each time.Experiment shows that the reverse position error of test point is within 0.06 mm.