Analysis of Visual Information Processing in Personnel's Raw Silk Testing and its Application to a Raw Silk Testing System

为了实现生丝检验的自动化，进行了本研究。去年开发了能够将所有缺陷分类为孤立缺陷的缺陷测试系统。然而，例如，两个或更多个次要缺陷相邻，出现的缺陷被分组，并且它们偶尔在实际测试中被判断为一个主要缺陷。为了解决这样的边缘提取问题，1998财政年度，我们研究了两个或两个以上相邻出现的缺陷的分类。结果，缺陷的边缘提取过程中，模糊理论的应用的计算机模型被构建的基础上的视觉信息处理的结果时，检查人员测试的缺陷的前一年。该模型采用模糊规则集，包括边缘提取规则部分和平坦部分提取规则部分。首先，根据边缘提取规则，利用模糊推理将所有出现的缺陷作为孤立的缺陷进行边缘提取。其次，基于相同的推理，将部分提取的缺陷分类为相邻缺陷。新系统的缺陷分类率能够通过这种方法提高约5%，与旧的一个。

To automate the raw silk test, this study was done. The defects test system which is able to classify all the defects as isolated ones has been developed in the previous year. However, two minor defects or more are for instance adjacent, the appearing defects is grouped, and they are judged one major defect in an actual test occasionally. To solve such edge extraction problem, 1998 fiscal year, we examined the classification of two or more adjacently appearing defects. As a result, the computer model of the edge extraction process of the defects by which the fuzzy theory was applied was constructed based on the results of the visual information process when the inspection personnel tested the defects by the previous year. The constructed model uses fuzzy rule sets, and has the edge extraction rule part and the flat part extraction rule part. The process of the system was explained as follows : first, the edges of all the appearing defects were extracted as isolated defects based on the edge extraction rules using fuzzy inference. Second, some of the extracted defects were classified as adjacent defects based on the flat part extraction rules using the same inference. The classification rate of the defects of the new system was able to be improved by this method by about 5%, compared with the old one.