论文部分内容阅读
本文介绍一个用于鉴别重叠工件的计算机视觉系统。系统包括预处理、模型训练、图象分割三部分。 为了完成图象分割,利用了边界与阴影的统计与结构信息,所提出的分割算法由两部分组成:曲线段的聚合,假设与验证。第一步能对各工件定位,定方向并能将属于同一工件的线段聚合在一起;第二步通过假设与验证过程鉴别最上面的工件。关于顶部工件的假设是基于第一步获得的信息,验证是通过搜索来寻找支持假设的有关事实。由阴影而形成的内边界,被用于验证所作假设。 数据库表示工件的结构和几何特性。三个框架B、I、M,分别表示外部边界,内边界和模型的特性,并组成一个三框架系统。 本文示出了两个与三个工件重叠的不同情况的实验结果。
This article describes a computer vision system for identifying overlapping workpieces. System includes pretreatment, model training, image segmentation in three parts. In order to complete the image segmentation, the statistical and structural information of the boundary and shadow are utilized. The proposed segmentation algorithm is composed of two parts: the aggregation, hypothesis and verification of curve segments. The first step is to position and orient each workpiece and to group together the segments that belong to the same workpiece. The second step is to identify the topmost workpiece through the Hypothesis and Validation process. The assumptions about the top artifacts are based on the information obtained in the first step, verified by searching to find relevant facts that support the hypothesis. The inner boundaries formed by shading are used to verify the assumptions made. The database shows the structure and geometry of the workpiece. The three frames B, I, M, represent the characteristics of the outer boundary, the inner boundary and the model, respectively, and form a three-frame system. This article shows the experimental results of two different situations that overlap with three workpieces.