Knowledge Representation about Images Based on Linguisitical and Numerical Concepts

1. 开发了一个基于lisp内核的面向对象图像处理系统，该系统对图像等数值数据支持特殊的数据结构record array。新实现了日语处理单元。这个系统可以整合各种各样的知识。关于图像和物体的知识以语言、数字和几何的方式表示。这些表示被实现为模型和它们自己的推理系统。将模型定义为抽象的模型类，并通过激活必要的图像处理方法构建模型推理系统，特别是利用遗传算法将三维模型与图像进行匹配，实现有效的三维推理。对于物体识别和图像理解等问题，我们所要做的就是定义一个继承抽象模型类的对象类。类的继承是对模型的自动集成，不需要逐个问题构建模型推理系统和图像处理系统。

1. An object oriented image processing system based on lisp kernel was developed, in which special data structure called record array was supported for numerical data like images. Japanese language processing unit was newly implemented. This system enables to integrate the various kinds of knowledges.2. Knowledges about images and objects were represented linguistically, numerically and geometrically. These kinds of representations were realized as models and their own reasoning systems.3. The models were defined as abstracted model classes, and their reasoning systems were constructed by some methods which activate necessary image processings, especially matching between 3-D model and images by genetic algorithms enabled effective 3-D reasonings.4. All we have to do for some problems such as object recognition and image understanding is to define an object class which inherits the abstracted model classes. The inheritance of the classes is to integrate the models automatically, and we do not have to construct neither model reasoning system nor image processing system for problem by problem.