Concave and Convex Area on Planar Curve for Shape Defect Detection and Recognition

Abstract

A shape representation based on concave and convex area along a closed curve is presented. Curvature estimation is done to the input curve and searched for its critical points. Splitting the critical points into concave and convex critical points, the concave and convex area is computed. This technique is tested on shape defect detection of starfruit and also to shape recognition. In the first case, defect is measured with concave energy and obtained a stable measure, which is proportional with the defect. In shape recognition, starfruit's stem is identified to remove it from the starfruit shape, as it will contribute to false computation in defect measurement