A robust Hough transform technique for description of multiple line segments in an image

The process of using the Hough transform (HT) to detect lines in an image involves the computation of the HT for the entire image, accumulating votes in an accumulator array and searching the array for peaks which hold information of potential lines present in the input image. The process of peak formation generates a butterfly shaped spread of votes in the accumulator array. The authors have used this property to adaptively define windows of interest around a detected peak to facilitate the description of multiple line segments within an image in terms of the coordinates of their end points. The developed technique has been employed to test several images composed of multiple line segments and the results in terms of accuracy of the determination of line segment mid points are presented. While most methods which employ the HT to detect line segments cannot handle the case of separate line segments formed by a colinear set of points, it is shown that the developed method can successfully do so. This algorithm would find applications in different areas of machine vision like robotics and manufacturing systems. Results of the application of the developed method, to detect lane markers and curve signs from a road scene captured by a CCD camera, to aid in the maneuvering of autonomous vehicles are presented.