| In engineering application and measurement experiment,displacement measurement is an indispensable link.From traditional contact measurement to non-contact measurement such as laser measurement,ultrasonic measurement and displacement sensor measurement,displacement measurement has been developing towards automation and high precision.With the continuous development of measurement technology,many current measurements methods are limited in some scenarios,and displacement measurement based on binocular stereo vision has begun to play a role.Current binocular vision systems are mostly aimed at the measurement of static objects in close range.The measurement of long-distance moving objects is not much,and there are some shortcomings such as low accuracy and poor real-time performance.Therefore,a binocular vision measurement system for measuring the depth and displacement of long-distance objects is designed in this paper.The research includes four aspects: binocular camera calibration,target detection and tracking,stereo matching of binocular images and target depth and displacement measurement.Binocular camera calibration: Through the research and comparison of calibration algorithms,the system calibrates the cameras useing Zhang's calibration method,and describes the calibration process in detail.Then,Zhang's calibration is realized through MATLAB calibration toolbox and opencv visual library function,and the results are compared and analyzed.Target detection and tracking in binocular vision system: The commonly used target detection algorithms are verified by experiments,and the advantages and disadvantages of the algorithm are analyzed.The Mixture Gauss Background Modeling method is used for target detection,and the voids and noises in the background image are filled and filtered.The KCF algorithm is used for target tracking.Inder to solve the problem of fixed scale of KCF algorithm,An adaptive KCF algorithm based on one-dimensional correlation filter and variable Gauss window is designed,and the effectiveness and performance of the algorithm are verified by experiments.Stereo matching of binocular vision: Through analyzing the common stereo matching algorithms,ORB algorithm is used to complete the stereo matching of the system.Scale pyramid is added to ORB algorithm to make it scale invariant;combined with improved KCF algorithm,the feature extraction area of the algorithm is reduced from the image to the target tracking frame area to improve the speed of feature point searching;FLANN algorithm is used to filter feature points,reduce the mismatch rate of the system and achieve real-time correct matching.Measurement of target depth and displacement in binocular vision: The depth of binocular vision is calculated by triangle similarity principle,the three-dimensional coordinates of the target are calculated by the depth obtained,and the displacement of the target is calculated by the transformation of the three-dimensional coordinates of the target.Two methods are proposed for depth calculation: the first one is to take the average value of feature matching points in left and right images after stereo matching,whose position is basically in the center of the target,and to calculate the depth using the average value of feature matching points in left and right images;The second method is to find the center of mass of the left and right image targets,and to find the depth of the object through the center of mass.The depth values obtained by two methods are used to calculate the three-dimensional coordinates,and then the displacement of the target is calculated.The experimental results are compared with the actual measurement results,and the error analysis is carried out.In the current experimental environment,the experimental results show that the research is reasonable and feasible.The experimental data obtained are compared with the theoretical data.The error is also within the allowable range.It shows that the system has certain accuracy,real-time and stability. |
PDF Full Text Request