An improved fitting algorithm for contour curve based on covariance matrices
Yanfang Li · Machinery Design and Manufacture · 2010
An improved fitting algorithm based on covariance matrices for planar contour curve which includes the features of line and arc in the field of industrial vision is proposed.Firstly,the contour data points which have been exacted by 8-connectivity method are ordered,and the smallest bounding box of contour curve is computed,which can be used to recognize the scope of feature points according to known curve type.Then,the region of support of each point is defined adaptively based on the eigenvalue analysis of its covariance matrices,which can be used to exact feature points.Finally,the contour curve is reconstructed by using least-squares curve fitting technique.The experimental results demonstrated that the proposed fitting algorithm for contour curve based on covariance matrices is effective and precision,which make an important function in online inspection.