Research On The Surface Profile Detection Of Large-scale Wind Turbine Blade Based On Close-range Photogrammetry

Close-range photogrammetry is a non-contact measurement method that can quickly and accurately obtain the spatial information of the object,so it can be widely used in full-scale detection of large and complex curved objects.This thesis uses a combination of analysis and research to conduct in-depth research on the key technologies in close-range photogrammetry,selects large-scale wind turbine blade as experimental object,and uses the close-range photogrammetry to detect surface shape in the middle section.The main research contents are as follows:(1)The target areas are coarsely located by using the connected domain analysis in the Blob algorithm to accurately and quickly extract shape features,then combine the least squares fitting algorithm to locate the sub-pixel-level center of the marker points.For the pseudo-marker points existing on the coding loop,this thesis proposes to use the area difference criterion to remove them.The experimental results show that the marker point detection method used in this thesis is feasible and the positioning accuracy meets the requirements.(2)The commonly used coordinate systems in close-range photogrammetry and their transformations,basic conditional equations and camera pose orientation algorithms,space forward intersection method,beam adjustment optimization algorithm,and basic theories of 3D reconstruction are studied.The relative orientation algorithm and the direct linear transformation solution are used to solve the camera’s external orientation elements.The initial three-dimensional coordinate values of the marker points are calculated based on the same-named image points,internal and external parameters of the two images combined with the least squares method.Finally,the point-by-point method of beam elimination method adjustment calculation is used to perform the overall global optimization iteration.The space forward intersection method is used to reconstruct the three-dimensional spatial coordinates of all marker points,and the single-camera three-dimensional reconstruction is completed.(3)The Nikon D7200 camera used in the experiment is a non-measurement camera and has a large lens distortion error.In this thesis,a ten-parameter distortion model is used to compensate the imaging distortion caused by the lens.The camera is calibrated by using the self-calibrating beam method adjustment in photogrammetry.Camera calibration experiments are performed using calibrated crosses,which are taken from different orientations.Based on the relative orientation algorithm,direct linear transformation solution,and self-calibrating beam method adjustments,we can obtain the camera internal,external orientation element values and distortion coefficient values.The calibration results are used to analyze the reprojection error of the mark points,and an accuracy analysis experiment is performed on the scale reconstruction.The experimental results show that the maximum average reprojection error of the image is 0.233 pixels,and the average measurement error of the scale is 0.051 mm,indicating that the self-calibration algorithm based on the beam method adjustment is feasible and has high accuracy.(4)The accuracy of matching non-encoding marker points with epipolar geometry constraints of multiple images is studied.In this thesis,multiple images are used for non-coding point matching.The experimental results show that the matching accuracy of 49 pairs of non-coding points is 100%,indicating that this method can accurately match non-coding points.And a small-scale scene reconstruction experiment is designed.The algorithm studied in this thesis is used to complete the three-dimensional reconstruction of two types of mark points.The results are compared with the XJTUDP system reconstruction results.Experimental data show that the reconstruction results in this thesis are accurate and meet the requirements.Based on the research of the above-mentioned theoretical algorithms,combined with the point cloud stitching algorithm in 3D reconstruction,photogrammetry technology is used to detect the surface shape of large-scale wind turbine blade,and achieve a good result.