A Fundamental Study on the Digital Recognition of Grinding Wheel
In this paper, the 3D morphology of a grinding wheel was modeled by the depth from focus. Firstly, the picture information of different heights was extracted by the up-down moving of the microscope. The operator Laplacian was adopted to distinguish the distinct and fuzzy areas in a picture. Then, the distinct image and height information was obtained. The information of height was distorted due to the occurrence of noise. In order to reconstruct 3D surface, a method based on Min/Max curvature flow was developed to remove noises. In the end, an abrasive grain in the image of a grinding wheel was segmented by the Mumford-Shah model. The results could be further developed to evaluate the worn status of grinding wheels. Introduction The examination of the wear of abrasive grain in the grinding wheel is very important for evaluation of performance of diamond grinding wheel. The three- dimensional (3D) reconstruction of grinding wheel topography can provide more information about wear of abrasive grains than common ways such as observation by optical microscope. Nowadays, there have been many techniques to be considered to obtain 3D data, for example, profilometry, the scanning electron microscope (SEM), the scanning laser microscope (SLM), stereo vision and so on. SEM is a powerful measuring tool, but the time needed for sample coating process and chamber air pumping is considerable. SLM is promising tool for 3D shape modeling, but still expensive for most of users. Stereo vision is simple and quick method to obtain the height information, but only the height of points which match in two corresponding images could been obtained. In this paper, a new method based on depth from focus
Depth from focus Level set Min/Max curvature flow Mumford-Shah model
Junfeng Gong Xipeng Xu
Provincial Key Research Lab for Stone Machining, Huaqiao University, Quanzhou, Fujian, China, 362021
国际会议
第14届全国磨粒技术学术会议(14th Conference of Abrasive Technology in China)
南京
英文
504-508
2007-10-26（万方平台首次上网日期，不代表论文的发表时间）