Research On The Correction Technology For Measurement Error Based On Binocular Vision

Under the background of the strategic policy of "Made in China2025",The computer vision technology has been widely used in real life,as an important branch of it,the binocular vision measurement technology has been developed rapidly.In binocular vision measurement,measurement errors generally come from two parts,one of which is due to the low calibration accuracy in the camera calibration process,the other is due to the low matching accuracy in the matching process.So,in the aspect of binocular vision measurement technology,the quality of camera calibration and image matching technology are especially important.Based on the research of binocular measurement vision technology at home and abroad,some researches were done in this paper on camera calibration and parameter optimization,distortion image correction and image matching,and the experimental simulation was carried on to verify the feasibility and accuracy of the research method.Finally,a binocular vision measurement method was proposed,which improved the measurement accuracy and corrected the measurement error.In order to improve the measurement accuracy and reduce the measurement error in this paper,in the research process of binocular vision measurement technology,the main research contents are as follows:(1)In order to improve the accuracy of camera calibration,an adaptive genetic simulated annealing algorithm 'based on the combination of the piecewise selection strategy and random sampling was proposed to optimize camera parameters.First of all,the initial value of camera parameters was obtained by zhang zhengyou's calibration algorithm,and then the parameters were nonlinearly optimized using the improved genetic-simulated annealing algorithm.Finally,the feasibility of the algorithm was verified by the simulation experiments.Compared with existing algorithms,it has been improved in terms of accuracy and operational efficiency.(2)In the aspect of image distortion correction,considering that the traditional correction algorithm needs to establish the mathematical model of distortion,and the distortion model can only approximately describe the relationship between the standard image and the distorted image,and the traditional correction algorithm has the disadvantage of large computation.Therefore,a BP neural network distortion correction algorithm(improved GA-SA-BP algorithm)is proposed in this paper based on the adaptive genetic simulated annealing algorithm,a combination of piecewise selection strategy and random sampling to correct the distortion images.This algorithm does not need to establish a mathematical model of distortion.At the same time,the improved genetic simulated annealing algorithm is used to optimize the weight and threshold of BP neural network,and makes up for the shortcomings of neural network that is easy to fall into local extremum.Compared with existing algorithms,this algorithm has improved the accuracy and reduced the number of iterations.(3)In the aspect of image matching,to improve the matching accuracy and time efficiency,an improved multi-scale Harris-SIFT matching algorithm is proposed in this paper.At first,the improved multi-scale Harris corner detection algorithm was used to extract the feature points in the image,and then the improved 36-D feature descriptor was used to describe the feature points,and once again the bidirectional matching strategy was used based on the Euclidean distance and the cosine similarity algorithm to match the feature points,Finally,the RANSAC algorithm was used to match accurately and eliminate the mismatched points.Experiments showed that the improved algorithm was enhanced in time efficiency and matching accuracy compared with existing algorithms.At the same time,in terms of measurement,firstly the three-dimensional coordinates of the point were calculated by matching the feature points in the left and right images,and then the distance was calculated by the distance formula between the points.Finally the error analysis was made with the real value to verify that the method proposed in this paper meets the precision requirements required for measurement.Compared with existing measurement methods,the measurement accuracy has been improved.