摘要
二维手绘形状识别是2D草图重建3D模型领域的关键课题。本文提出了一种二维手绘形状在线识别方法:通过二维输入设备采样获得的手绘形状数据为离散的点,这些数据只有当被以某种数学形式表示时有更多的实际意义,Spline即是一种常用的表示形式。本文对手绘形状数据进行了参数化,在此基础上进行了Spline拟合。然而将一个手绘形状中的全部点序列进行Spline拟合时会使尖状角点被平滑,因而使手绘形状的重要特征发生偏离。为此,本文对一次拟合的Spline进行曲率计算,求取曲率的极大值,确定候选角点,去除伪角点最终确定角点.,从而以角点对采样点进行分段,然后对分段后的采样点进行二次Spline拟合,此过程中由直线和圆弧本身曲率的特点更增加了对于直线和圆弧的判定替代.实验结果表明,这种方法在获得更加接近手绘者意图的平滑效果较好的手绘形状的同时,保留了重要的角点特征。
That identification of two-dimensional Freehand shapes is a key fields of that 2D draft rebuilding 3D model. This paper proposes an identification method of two-dimensional Freehand shapes：Freehand shapes acquired from a 2D input device are represented in discrete coordinates. These data make more sense when they are represented in some mathematical forms,for which Spline is a due candidate. This paper parameterizes the original freehand shape data first,and then fit them using Spline. However,we have found that whole shape Spline fitting smoothes corner points toward circles,diminishing corners which are significant shape features. To remedy this drawback,we calculate curvature along this first pass Spline,search for the local maximums which represent corners,divide the shape data into segments by these corners,and finally fit each segment with Spline. Experimental results show that the proposed approach well retains shape corners and while smoothes segments between these corners.
出处
《微计算机信息》
2009年第34期117-119,共3页
Control & Automation
