| Cylindrical objects play an important role in representing the geometry of mechanical structure.The rotating parts represented by cylindrical objects are widely used in industrial activities such as mechanical manufacturing.Especially in the environment of automatic assembly,in order to achieve precise assembly and ensure product quality,it is important to detect the spatial state of cylindrical objects quickly and accurately.However,cylindrical objects generally have the characteristics of less surface texture or no surface texture,and are mainly made of metal or composite materials with high reflective characteristics and other factors,which limit the precise detection of spatial state for cylindrical objects.Due to the advantages of non-contact,high-speed and high-precision,3D measurement based on fringe projection measurement technology has been widely used in industrial detection,quality inspection and other aspects,and has become a research hotspot in recent years.In this paper,taking the spatial state detection of high dynamic range cylindrical object as the research object,aiming at the measurement difficulties of less texture features and surface reflection,an adaptive binocular fringe projection method is proposed by integrating binocular vision and fringe projection measurement profilometry,and its key technologies are studied,focusing on the problem of high-precision camera calibration,high dynamic range fringe image,fringe image filtering and cylinder fitting with incomplete point cloud.The main research contents include:(1)The extraction method of high robust calibration target center was studied.The calibration target image is the basis of camera calibration.The sub-pixel high-precision segmentation and ellipse center extraction of the calibration target image with circular feature were studied.Based on the theory of fuzzy quotient space,the dynamic granularity matrix space model for image segmentation was proposed,and its properties were described.The problem of image segmentation is described as the granularity conversion and jumping of the image at different levels,and the optimal fuzzy clustering population is obtained by information entropy method;quantum inspired group leadership optimization algorithm is proposed to solve the above mentioned optimal clustering initial population,so as to obtain the global optimal solution,namely segmentation threshold.Combining with the threshold segmentation method,the region of interest of each calibration target image can be segniented;a robust ellipse center extraction method based on the dynamic granularity matrix space model was also proposed,whose idea is to transform the ellipse center extraction problem into the optimal clustering problem of the ellipse center cluster.Firstly,the iterative least square method was used to get the ellipse center cluster.Secondly,the dynamic fuzzy cluster was used to get the optimal cluster based on the dynamic granularity matrix space model.Then,the multi-centers of the optimal cluster were compared with the least square fitting center to get the robust ellipse center.(2)The fringe image method of high dynamic range object was analyzed deeply.Aiming at the problems of high reflection and local specular reflection,an adaptive binocular fringe projection measurement method with high dynamic range was proposed.Firstly,through the system mathematical model,the nonlinear image intensity response function of the projector-objcct-camera was established,and the influence of the surface reflection characteristics of the object on the projection intensity was also analyzed.Secondly,the reflection characteristics of the surface pixels and neighborhood pixels of the object were obtained by projecting the uniform gray-scale images,and the optimal projection gray-scale value was calculated.Then,the modified gray values were mapped to the gray value of the corresponding point of the projection fringe through the binocular conversion relationship and the camera-projector conversion relationship.(3)The direction denoising method of phase-shifting fringe image was applied.Due to the influence of environment and electromagnetic equipment,the fringe image is inevitably polluted by noise,especially Gaussian noise.Firstly,the influence of Gaussian noise on phase calculation is analyzed,and it is deduced that the phase error caused by Gaussian noise also presents Gaussian distribution.In order to reduce the phase error caused by Gaussian noise,a fringe image denoising method based on the fringe phase-shifting field was proposed,which transforms the traditional single image independent filtering into the full phase-shifting image direction filtering.Secondly,the filtering direction of the full phase-shifting image was modified according to the fuzzy quotient space.Then,a functional filtering method with loyalty term was proposed,which is improved by the fourth-order direction partial differential equations.The fourth order directional PDEs method has faster convergence speed and more details of the original image.By filtering the four step phase-shifting fringe images,the phase errors caused by Gaussian noise is obviously reduced,and under different noise intensities,the method achieves better signal-to-noise ratio,which shows that it has higher robustness.(4)The cylinder fitting method with incomplete point cloud was proposed.When the binocular fringe projection system was used to measure the three-dimensional cylindrical object,the point cloud data is incomplete due to the noise,uneven sampling density,occlusion,lack of semantic relationship between samples and other problems,only part of the point cloud information,that is,incomplete point cloud,can be obtained.And the point cloud data is usually contaminated by abnormal points.Most of the existing cylinder fitting methods perform well on complete point cloud,but they are not robust for incomplete point cloud.Firstly,the mathematical model of point cloud and its errors are analyzed,then outliers and noise points were removed by bilateral filtering;then,the edge points on the cutting plane were used to fit the ellipse by random sampling consistency algorithm,and the cylinder axis was also fitted by the center point cluster of ellipses,so as to obtain the spatial state parameters of cylindrical object.In this paper,focusing on the spatial state detection of cylindrical object,taking the binocular fringe projection as the measurement means,the key technologies such as the high robust calibration target center extraction method,the high dynamic range adaptive binocular fringe projection measurement method,the fringe image denoising method based on the fringe phase-shifting field model and the incomplete point cloud cylinder fitting method based on cutting plane are studied.Through the measurement experiments of cylindrical objects with the diameters of 199.750mm,239.741mm and 276.299mm respectively,the average fitting error of diameter were less than 0.03mm,and the average deviation of axis vector were less than 0.02rad,which realizes the accurate measurement of spatial state for high dynamic range cylindrical object.The key technologies proposed in this paper are not only applicable to cylindrical objects,but also to other rotating objects. |
PDF Full Text Request