Research On Three-dimensional Shape Recovery Method Of Workpiece Based On Two-dimensional Images

With the development of technology in the field of machining,the analysis of the surface quality of workpieces is becoming a hot topic of research.Compared to two-dimensional analysis,three-dimensional analysis provides more visual and comprehensive surface information.At the same time,compared to traditional offline measurement method,in-situ measurement is increasingly in demand as it can provide timely feedback,improve machining quality and reduce costs.However,existing complex and expensive advanced 3D measurement means have cumbersome measurement processes,large equipment size and certain requirements for the inspection environment,which are difficult to apply in general machining environments and cannot meet the in-situ measurement needs.Therefore,finding a simple,fast and effective 3D recovery method that can be adapted in general machining environment has important research significance for the machining field.This paper firstly classifies and compares the technical means of 3D shape measurement,and selects the Shape from Shading(SFS)method and the photometric stereo method to study a simple and effective method to recover the 3D shape of the workpiece surface using only 2D images.Through theoretical derivation,algorithm writing,verification test and 3D surface recovery comparison test,a 3D shape recovery method based on the photometric stereo method of near-field point light source model is found,and the 3D surface measurement of the workpiece is realized.The specific research contents are as follows.(1)The theoretical derivation and implementation of the SFS linearization method and the photometric stereo method algorithm were carried out.The algorithms for both were written using MATLAB software,and the algorithms take the intensity of the 2D image as input and the surface depth as output.The basic processes are as follows: the Pentland method in the SFS linearization method uses linearization of the gradient term in the reflectance map equation,while the Tsai-Shah method directly discretizes the height and linearizes the Taylor expansion to finally solve for the depth;the photometric stereo method first calibrates the light source with a high light sphere,then uses the light source direction information to calculate the normal vector and finally solves for the depth.The two algorithms are then validated by virtual images and standard dataset images,respectively.The results show that the Pentland method produces overall saddle distortion,while the Tsai-Shah method and the photometric stereo method are more effective in recovering the whole object and its details in 3D.(2)The Tsai-Shah method and the photometric stereo method of a near-field point source model,were used to carry out comparative tests on the three-dimensional recovery of workpiece surface topography.The Tsai-Shah method and the photometric stereo method,which introduces a near-field point source model,were used to carry out comparative tests on the three-dimensional recovery of workpiece surface topography.The results show that the photometric stereo method has better texture,peak-to-valley agreement and less relative error in the 2D and 3D parameters for the 3D recovery of Si C workpiece surfaces.The photometric stereo method was then used to perform 3D recovery tests on workpieces with non-Lambertian surfaces,including a grinding silicon wafer,a 3D printed workpiece,a diamond grinding head and a cubic boron nitride(CBN)grinding head.The results show that the photometric stereo method based on the Lambertian hypothesis in this paper has more distortions and unsatisfactory results for the 3D recovery of non-Lambertian surfaces.(3)A hardware and software platform was developed for the measurement of the threedimensional shape of workpiece surfaces.The hardware platform includes a light source,a digital microscope and a computer,while a light source direction limiting structure is designed to replace the photometric stereo method light source calibration process and improve the measurement efficiency.The GUI interface of the software platform was developed using MATLAB software,which can realise the functions of image acquisition,image pre-processing,shape reconstruction and parameter calculation,etc.It has good human-computer interaction capability and is convenient for operators to use.