Research On Three-dimensional Shape Measurement Method Of Blade Based On Multi-view Stereo Vision

As the core of the airplane,the engine’s manufacturing and design quality can provide insight into the national level of industrial development and technological advancement.The most crucial portion of an aviation engine is the blade,and the performance of the entire engine is directly impacted by the accuracy of the blade part.The measuring results regarding the blade shape serve as an essential basis for determining the quality and lifespan of the blade during its production and use.If the blade quality is subpar,breakage,fracture,and other issues during aero-engine operation will result in a serious safety incident.Therefore,it is necessary to assess the blade shape in three dimensions accurately and effectively.The aero-engine blade serves as the research subject for this topic to study the blade 3D morphology measurement method based on multi-camera stereo vision,through the multicamera from four viewpoints to obtain the object 360 degrees of information in horizontal direction,and to achieve high-precision 3D measurement of the blade,with a focus on resolving the issues of discontinuous area error,multi-viewpoint point cloud fusion,and blade leaf shape parameter measurement in the structured light system.The article’s research is organized into four key sections as follows.(1)Overall system scheme design.The article firstly analyzes the technical requirements of aero-engine blade 3D measurement in detail,and designs a blade 3D morphology measurement scheme based on multi-camera stereo vision for the problem of reconstruction blind area in monocular structured light system,followed by a detailed introduction of the system composition,design of the system structure of four cameras and single projector,and hardware selection and system construction.The hardware parallel control timing is constructed in accordance with the characteristics of simultaneous acquisition and reconstruction of multiple eyes,followed by the design of the overall algorithm flow and software architecture.(2)Structured light 3D reconstruction method based on discontinuous region error correction.Aiming at the discontinuous region error problem caused by the free complex surface of the blade,the discontinuous region phase error correction is integrated into the system calibration,and the 3D reconstruction method based on the discontinuous region phase error correction is designed.Firstly,the camera calibration principle is clarified in detail,and then the camera calibration is completed.Secondly,for the phase error problem in the discontinuous region,the principle of error generation is analyzed,mathematical modeling is performed,and the phase error correction method in the discontinuous region is proposed to complete the correction of the phase in the discontinuous region.The inverse camera method is then utilized to finish the projector calibration based on the corrected phase.Finally,the system calibration is performed utilizing the binocular calibration principle in order to calculate the three-dimensional point cloud of the blade from the two-dimensional stripe image.(3)Multi-camera stereo vision based on point cloud alignment and attitude correction of the blade.A shadow removal approach based on temporal gray scale is proposed employing the phase shift principle to address the shadow issues brought on by the notable blade distortion.This method effectively eliminates the shadow in the blade’s streak image and lowers the noise of the reconstructed point cloud.A point cloud coarse alignment approach based on the parameters of the multi-camera stereo vision system is proposed for the blind area problem of the monocular structured light system for completing the point cloud information in the blind area.In order to tackle the problem of missing point clouds caused by overexposure of the blade surface,researchers suggest a method to fix the reflective area based on point cloud fine alignment,and generate a 3D point cloud with the overall shape of the blade.To address the issue of difficulty in setting the blade measurement reference,researchers present an attitude correction method based on cylindrical surface recognition to correct the blade attitude.Lastly,experiments are designed to validate the accuracy of the proposed point cloud alignment and attitude correction methods in this study.(4)Blade precision measurement method based on contour fitting.Firstly,for the problem of uneven point cloud density of the blade,the point cloud pre-processing method based on voxel sampling is proposed to make the point cloud density of each region of the blade surface consistent.Secondly,a shielded Poisson reconstruction approach based on normal constraints is presented to eliminate the redundant information of object edges after surface reconstruction,reduce the wrong hole filling,and improve the surface quality of the blade model after surface reconstruction to address the problem of pseudo-edge and wrong hole filling of blade surface edges.Secondly,the parameter definition of the leaf model is studied in detail,and the method of slicing the leaf model based on the open-source library VTK is proposed to obtain the crosssectional information of the leaf.Finally,a polynomial fitting-based blade parameter modeling method is proposed to complete the measurement of blade leading and trailing edge circular curves,leaf basin leaf back curves and maximum blade thickness.This project mainly researches the multi-camera stereo vision 3D reconstruction algorithm for aero-engine blades,and develops the blade morphology 3D measurement system program based on it,which mainly includes the system stereo calibration,3D reconstruction,point cloud processing and 3D measurement and other functional modules.In order to verify the effectiveness of the proposed 3D measurement algorithm based on multi-vision stereo vision,the calibration and 3D reconstruction accuracy test,point cloud alignment and surface reconstruction test,blade model slicing and cross-section measurement test were conducted based on a blade.The test results show that the designed and built multi-camera stereo vision3 D measurement system has high accuracy and robustness,the absolute accuracy of 3D reconstruction is 0.03 mm,the repeat accuracy is 0.01mm;the absolute accuracy of blade model measurement is 0.03 mm,which meets the requirements of the subject.