This paper explores a concept of informal user interface for graphic computing, which can be characterized by two properties: stroke-based input and perceptual processing of strokes. A prototype system, named Magic-Sketch, is also outlined, which is well suited to natural and creative processes by making abstract problems more concrete informally. Two important issues to capture users' intention effectively for implementing sketch-based graphic input system, adaptive stroke segmentation and incremental sketch recognition, are discussed in details.

CAD系统对方案设计支持的关键是建立独立于几何细节而又表达功构关系的装配模型。本文试图运用逻辑装配单元的定义来实现产品功能需求和结构表现间的有效关联，分别讨论了方案设计中功能描述和几何构形表示方式，形式化表达了方案设计的逻辑造型和装配建模过程，给出了支持自顶向下方案设计的装配模板定义，为方案设计的计算机辅助实现作了有益的尝试。

本文阐述了一种面向概念设计的逻辑结构造型方法，提出了独立于几何细节而又表达功构关系的逻辑结构特征定义和表示方式，给出了支持自顶向下概念设计的装配模板定义，并采用形式化表示形式，初步建立了计算机辅助概念设计理论。文中所提出的方法为计算机辅助概念设计系统的实现提供了初步框架。

手绘图形输入是表达设计意图，实现抽象概念与具体图形间相互转换的自然而有效手段之一。本文提出并实现了一个面向构思的手绘草图输入解决方案。该方案具有三个特点：一是将用户输入的笔划经过图形预处理、特征识别、图形拟合和规整等过程，自动地预测用户的基本图形构造意图；二是将不同复杂层次草图图形元素构成抽象为空间关系图，并在图匹配计算过程中引入约束部分枚举方法智能预测匹配的有效状态，可适应不同复杂度图形识别的要求；三是采用支撑向量机主动式增量学习方法主动“分析”用户增量历史数据而有效鉴别和快速识别用户手绘笔划特征，运用增量决策树记录草图的笔划构成及其手绘过程来捕捉用户草图的手绘习惯，进而利用模糊匹配在草图绘制过程中预测和识别复杂图形，初步解决了在线草图识别中用户适应性问题。

Sketching is a natural and effective way to capture designer's brainstorms ideas. This paper presents a sketch-based graphics input system for conceptual design (SketchGIS), which is mainly based on online graphic recognition and dynamic user modeling. Online graphic recognition can discover latent primitive shapes among user-drawn strokes and show the regularized shape on the screen immediately based on spatial-relation graph (SRG) representation and conditioned partial matching by shape preparing, feature recognition, shape filtering and regularization. Dynamic user modeling method is used to recognize and predict sketchy composite shapes before they are completely drawn for user adaptation, which is based on the SVM-based incremental active learning method and the incremental decision tree induction method. Experiments show that the proposed method in this paper can yield good recognition precision, and the prototype system proposed provides fine user interaction effect, especially for UML diagramming.