Three-Dimensional Displacement Measurement Based On Monocular Vision

Monocular vision measurement technology has the advantages of non-contact,simple structure,strong reliability,and high precision.It has already played an important role in the fields of aerospace,civil engineering,industrial production,robot navigation and target tracking.In the field of industrial production,three-dimensional displacement(deformation)is often used as an evaluation of mechanical structure performance and service life;it makes three-dimensional displacement measurement one of the significant steps during the structural condition monitoring.The traditional displacement measurement methods require manual participation,which are time-consuming,laborious,and difficult to be digitized.Considering those problems,this thesis designs a real-time measurement system for three-dimensional displacement based on monocular vision and intends to use it for expansion measurement of power station equipment.The simulation experiment proved the accuracy of the proposed WIEPn P algorithm.And the actual prototype experiment shows that the three-dimensional displacement measurement system designed in this thesis has obtained a measurement error of less than 1mm in the x,y,and z directions,respectively.It is feasible and practical in field applications.The camera pose measurement needs to match the world coordinates of n pairs of points with their pixel coordinates in real time.Firstly,it’s necessary to design a cooperative target in the real-time displacement measurement system.This thesis designs a cooperation target using circular markers,and determines the number of landmark points to be 6.The landmark points can be coplanar through the simulation and comparison experiment of the Pn P algorithm.Combined with the characteristics of the backlight source and the settable exposure time of the camera,the system can effectively reduce the influence of the image background when sampling the image in the low-exposure scene.In this thesis,an ellipse detection algorithm is used to extract the pixel coordinates of the circle centers,and then the world coordinates of the circle centers are matched according to the order of the pixel area of the circulars.The calibration plate image and target image needs to be denoised and enhanced to improve the accuracy of camera calibration and 3D displacement measurement.This thesis did comparative studies on image enhancement algorithms and image denoising algorithms,and then introduced the principle of these algorithms and their application scenarios.Eventually,the differences among the images with different preprocessing algorithms were compared and analyzed.As for the camera calibration,the checkerboard images needed to be undistorted first,and then histogram equalization and filtering are needed to improve the accuracy of corner detection.Before the image mark point recognition and detection,the image will be undistorted,and then the contrast stretching image enhancement method,the median filtering and the gaussian filtering method will be applied,which can not only ensure the enhancement of the effective information of the images,but also protect the pixel information.The key of the monocular vision displacement measurement system designed in this thesis is to solve the camera’s pose,that is the relative positional relationship between the camera and the measured target.Real-time displacement measurement requires a fast pose estimation algorithm.In this thesis,an analytical EPn P algorithm is selected.Based on this algorithm,an appropriate weight coefficient is set for the error of each marker point to reduce the influence of the spatial distribution(such as the points depth changes)of the marker points on the algorithm results.After the initial result is obtained,several iterations will be performed to suppress the influence of image noise and abnormal point data on the pose calculation.The simulation experiment results show that the real-time performance and accuracy of the proposed WIEPn P algorithm meets the requirements of the measurement system.In the end,the displacement measurement system designed in this thesis is verified in prototype experiments.The experimental results show that the three-direction errors of the system are less than 1mm and the time required for each measurement is less than 2s,which can meet practical application requirements.