Line and circle finding by the weighted Mahalanobis distancetransform and extended Kalman filtering

摘要

The paper presents a new parameter space approach, called theWeighted Mahalanobis Distance Hough Transform (WMDHT) whose main meritis to incorporate formal stochastic image and feature noise models. Itis aimed at improving the efficiency, accuracy and reducing the size ofthe accumulator arrays by combining it with extended Kalman filterrefinement. It works by detecting image feature points in theneighbourhood of a contour instead of exactly on the contour through aMahalanobis distance measure modified by a weight function inverselyproportional to the distance between the point and an ideal contour. Themethod is applicable to geometric features of any dimensionality and thepaper illustrates it by considering detection of straight and circularsegments