手绘 CAD 图形输入的识别

Recognition of Freehand Drawing in CAD

讨论了手绘ＣＡＤ图形输入的识别问题，提出了手绘图形识别方法，对曲线进行线性逼近以抑制噪声并减少识别数据量，提取几何与非几何特征进行粗分段，在此基础上利用基于ＢＰ算法的神经网络分类器进行直线和圆弧的分离和识别，以正确实现分段识别并重构图形。

The authors deal with the recognition of on line freehand drawing in CAD. It uses a linear approximation technique based on area deviation to eliminate noise and other criteria based on the intrinsic geometry features of curves (such as curveture, angle span etc.) and non geometry feature (drawing speed) for segmentation and recognition. The turning points were checked first and sub curves were divided, then the straight lines and arcs were segmented depended on the curve′s features by a neural network classifier.