Research On SFF 3D Reconstruction Technology And System Development Based On Microscopic Sequence Image Set

With the widespread use of micromechanical systems in industrial,military,medical,and other fields,the demand for detection of micro mechanical parts and their processing tools has increased significantly.Since the observation of these micro-sized structures is difficult using conventional means,it is important to study the observation techniques and equipment for fine and micro structures to promote the development of micro-fabrication technology.Based on SFF method in this paper,the cutting part of the micro-diameter two-blade end mill is taken as the object,and the local-focus sequence image set is obtained through the developed microscopic auto-observation device,so as to realize the three-dimensional reconstruction of the object.The specific work completed is as follows:Microscopic auto-measuring device for automatic image acquisition and observation of micro parts is designed.According to the observation requirements,the structural design and electrical control system design of the microscopic auto-measuring device are completed,and the construction of the assembly and the actual system is realized.In order to realize the automatic image acquisition of the object,an automatic locating algorithm for the start of the local-focus sequence image acquisition in SFF is proposed.The algorithm adopts a phased search strategy.Firstly,the large-step traversal search of full-stroke is carried out based on the rate of change of image gray value,which realizes efficient locating in the fast search phase of the algorithm.Then,based on the improved Tenengrad sharpness evaluation function,the small-range small-step accurate search is carried out,which realizes the final high-precision positioning of the acquisition start of the local-focus sequence image.The above search algorithm is implemented and integrated with the secondary development program of image acquisition equipment and PMAC multi-axis motion control card,and the development of microscopic auto-observation device control software is completed.According to the principle of SFF method,the single view 3D surface reconstruction is performed on the object.Through the analysis of the imaging system comprehensive depth of field,the acquisition step of the single-view sequence image is determined,and the local-focus sequence image set of the single-view direction of the object is acquired.It is proposed to pre-screen all the pixels of the full sequence set to eliminate the pixels that are difficult to obtain the trusted depth values,the effectiveness of the fitting solution process during extracting image depth values is improved.A method for correcting and missing point filling of the extracted image depth value data based on the R2 determination coefficient of the pixel in the depth value solution is proposed.This method can effectively improve the accuracy of the depth value obtained from the image.The pixel image coordinates and the acquired corresponding depth value data are converted into point clouds,and the single-view surface is reconstructed by using triangular pieces and NURBS pieces,which all achieve good result.The three-dimensional reconstruction of the subject is performed using the acquired multi-view point cloud data.The different views transformation relationship of point cloud data coordinate system in three-dimensional space is determined.A two-step registration method is proposed,which uses the rotation angle of view during image acquisition as basis of coarse registration and then uses ICP algorithm as the way of accurate registration,in order to avoid the disadvantage of easy to fall into local optimum during ICP algorithm registration,and the point cloud data mosaic of multi views in the unified reference coordinate system is realized.Subsequently,the jointed point cloud data is reduced by K-means clustering algorithm,and the multi-view 3D model is reconstructed by using triangular pieces.Using the reconstructed micro-diameter end mill cutting part model,the second back angle of the bottom edge is measured by the cross-section measurement method and the spatial plane fitting measurement method.