在這篇論文中,我們提出了一種相當特殊,且具有高效率的方法,利用輪
廓的鏈碼所具有的方向性來計算弧線段曲度的估計值,進而決定一個物件
輪廓上的臨界點。然後求出以每個臨界點為頂點的弧線段的角度,來做為
這個物件的特徵,再使用修改過的WLD來計算物件之間的距離,以求取其相
似的程度並將物件加以分類。經過修改後的WLD具有比對模糊性,能夠對雜
訊有更高的容忍度,進而使物件之形狀辨識及分類更具彈性。

In this paper, an approach to shape recognition/
classification based on Di-rection Flow Change (DFC) function
and extended WLD method is presented. The DFC function is able
to locate a set of critical points for a shape. A sequence of
angles corresponding to these critical points can be computed to
serve as the major features of a shape image. The extended WLD
method measures the distance between two arbitary shapes whose
representation are sequences ofangles. Using DFC function to
find critical points is very efficient. Shaperecognition/
classification becomes very flexible of the extended WLD
methodis both effective and robust for planar shape recognition/
classification.