In this paper, we present a novel method that is capable of detecting multiple shapes, including: straight line, circle, ellipse, triangle, rectangle and other polygons. Multiple shapes' detection is realized via analysis of edge orientation curve, which are obtained by tracing the contours of pre-segmented image region, and recording contour pixels' orientations. A modified histogram is generated from the edge orientation curve, for initial shape guess, by two groups: line/polygon, and ellipse/ellipse arc. Based on the primary shape recognition, target objects are precisely estimated through detecting the corresponding equation's parameters.
展开▼
