The Application Research Of Image-Based Modeling In Reverse Engineering

Currently, the 3D scanner based on grating projection method has been widely used in the field of reverse engineering, since it has advantages as high measurement precision and strong system stability. However, it also exhibits some defects such as complicated operation process and small measurement range due to the limitation of measurement principle. Simultaneously, image-based modeling (IBM) with many practical advantages as lower cost, virtual reality and easy operation is acquiring of the sequenced images, analyzing and matching the images’ characterization, which has been widely applied in 3D movie, virtual reality,3D games, et al. Nevertheless, the IBM cannot be directly used in the above reverse engineering fields where there are high precision requirements. Therefore, there is an important practical significance to search for the method of improving the modeling accuracy and apply the IBM to the fields of reverse engineering.To improve the modeling accuracy, a revised IBM method with nondestructive measurement capability has been presented based on the image sequence acquisition. Firstly, the image enhancement method include Gaussian filtering, median filtering and histogram equalization were applied in the image processing. Secondly, through the analysis of the characterization of the sequenced images the sparse point cloud is obtained by self-calibration treatment for the processed sequenced images in the PhotoScan. To reduce the impact of lens distortion on acquisition accuracy, the Zhang Zhengyou calibration algorithm was used to obtain the lens distortion coefficients. Finally, the 3D model has been built after the process of PMVS Dense treatment and triangulation on sparse point cloud.By analyzing the difference of the images through 3D modeling on static and dynamic objects, two kinds of image sequence acquisition devices have been presented by using an array camera. Based on the analysis of 3D modeling principle of IBM, systematic experiments have been implemented to study the factors of modeling accuracy, such as the image pretreatment method, lens distortion, camera parameters, camera array layout, surface texture and structure characteristics of object. Experimental results show that the image pretreatment method is an essential step, which significantly improves the model quality. Eventually, an optimal system is built by the optimization of the system parameters. The reconstruction accuracy of the optimal system is obviously better than that of the traditional 3D scanner based on referencing to the European Union measurement testing standard (VDIVDE2634).