Research On Image Feature Extraction And Its Aplication In Reverse Engineering3D Modeling

Currently, Reverse Engineering is widely used in new product’s design andmanufacture. Image feature extraction is one significant part of Reverse Engineeringresearch. It is the object of study of three dimentional modeling in Reverse Engineering.The feature lines are extracted from pictures and three triangular meshes.This studyproposes two method. The first one is creating holes based on image feature extraction.The another one is an algorithm for extracting feature curves on triangle meshes basedon multi-seed points.In allusion to hole modeling in Reverse Engineering for improving positionalprecision, the study has proposed a method of creating holes based on image featureextraction. The process includes three main steps. First, the images should be handledwith image enhancement tecnology. Next, extract the contour lines of holes withimproved morphological gradient operation. Finally, the extracted feature lines areimported in modeling soft. The method has been applied to one car lamp holder. Thecomparative analysis of the hole position precision between traditional modeling waysand the proposed method have been done. The results show that this method is moreaccurate than traditional ways toward hole modeling.For extracting feature lines in triangle meshes, this study also proposes analgorithm for extracting feature curves on triangle meshes based on multi-seed points.Feature line positions of triangular meshes were detected on the basis of curvature valueand normal vector, highlighted these positions with different colours. To assign severalseed points on the proper position, and compute their feature attributes. Construct thenode assessment functions and choose correct path nodes. Finally, smooth the curvesextracted from the triangular meshes. Experimental results show that the method couldbe effective not only in extracting closed feature curves, but also opened featurecurves and the region features are changing gently. It is higher accurate than othermethods to detect feature at the branch feature regions.