Detection of linear geological features (jointing) by Hough Transform of multispectral remotely sensed images

摘要

This paper discusses the recognition of geological lineaments seen in multispectral images of the Earth collected at optical wavelengths. The possibility of discerning jointing of rocks from natural and artificial image objects is thoroughly discussed. This problem is addressed by using an original image classification algorithm that enables us to detect urban areas and rivers. Once that a reliable sub image freed from possible artefact causes is isolated the true analysis of linear features is performed. This analysis takes advantage from the use of the Hough transform, in order to carry out the automatic identification of linear features and their analysis. The performance of the algorithm has been investigated by processing high resolution aerial photogrammetry and Thematic Mapper images. Tests as far executed have shown a good ability of the algorithm to accurately map image spatial features with linear morphology, and very rare occurrence of mix-up with different image objects.