Studies on expert system of tree distingushed for education using photo-database

在过去的两年中，我们研究了树木识别专家系统。本文的研究目标是出现一种可树区分的知识，并构建原型专家系统。针对利用相机、视频和数码相机采集树木的照片数据，提出了整理和维护照片数据库的方法。利用pc编写了一些照片数据库的维护和逆向程序。接下来，我们利用专家壳构建了两个针对北海道京都大学森林64种树种的专家系统。推导的方法是比较流行的IF-THEN法。我们根据每棵树的具体情况安排分类指标，如叶子的形状、叶脉的形式等，并构建知识库。对推导路径进行分析，发现所有分类指标都没有必要进行区分。第一个专家系统类似于植物书籍的参考。这是一种利用分类指标扩展到分支演绎的方法。叶片的长度和宽度采用富士演绎技术。第二个专家系统与第一个专家系统相同，直到中间分类。中间分类完成后，将所有树的照片和分类索引显示在显示屏上。如果用户选择其中一个与对象树的分类索引相对应，则显示树的特征。对于不了解分类索引术语含义的初学者来说，这种方法对于学习如何区分树的名称非常有用。

We studied on expert system of tree distinguished during past two years. The objects of this studies are to appear a knowledge for tree distinguished and to construct of prototype expert system. We get some methods according to how to gather photo-data of tree using camera, video and digital camera, to arrange them and to maintain photo-database. We make some programs of photo-database for maintenance and reverse using P.C..Next, we construct two expert systems of tree distinguished against 64 species on a Kyoto University forest in Hokkaido using expert shell. The method of deduction is IF-THEN method that is popular. We arranged classify indexes according to specifics of each tree which are a shape of leaf, a form of leaf vein and etc.and constructed knowledge database. Analyzing for path of deduction, it was appear that all classify indexes are not necessary to distinguish. The first expert system is similar to reference of a plant book. That is a spreading into branch deduction method using some classify indexes. The technic of Fuji deduction is used for length and width of leaf. The second expert system is same to the first until middle classify. After middle classify, all proposed trees are shown at display with their photograph and classify indexes. If user selects one of them that is corresponding to classify indexes of object tree, the tree characteristic is shown. For a beginner who doesn't understand a mean of classify index's term, this method is useful to learn how to distinguish tree names.