Extraction of linear features in multispectral imagery

Abstract

Linear feature detection is a classical problem in road extraction. Many techniques have been developed over the past years to extract salient road networks from aerial or satellite imageries. Such techniques include morphological methods, like thinning, differential geometric methods, like ridge-and-valley detection, active contour tracing intensity contrast and others. Most of these techniques report very good results on selected domains of grey level images. In contrast, much less effort has been devoted to exploit the usefulness of colour information for linear feature extraction. In this paper, we explore the use of colour to perform linear feature extraction. The scope of our study focuses on the extraction of linear feature points. The connection of feature points is in the process of development.