Realtime circular object detection based on the statistical compactness estimation [İstati·sti·ksel tikizlik kesti·ri·mi·ne dayali gerçek zamanli dai·resel nesne tespi·ti·]

Abstract

Gathering statistical and geometrical information by processing the shape contours is the common way of feature extraction on object detection and recognition studies. Compactness is an important shape descriptor which specifies the similarity between a shape and a circle. In this study, we propose a new compactness measure based on examining the distribution of the contour moments with respect to the shape's centroid. First, the contours are extracted with the Edge Drawing algorithm from the image. Then, the contours moments are computed and their distributions are examined. As a result, detection of the circular shapes among the extracted closed contours with a desired circular similarity becomes possible. With its high accuracy and low complexity, the proposed method is a convenient for realtime applications