In this paper

a new technique to the problem of road recognition in TM images is presented. The algorithm takes advantage of the characteristics of TM images. It uses steerable filters to collect candidate road objects in the images

then the chain code is used to represent these candidates and computer their parameters

and in the last step we recognize those true roads according to some knowledge and make the result more rational by post - processing. The result is satisfying considering the low resolution of the TM images. Results on the TM images of Taiwan area are given to demonstrate the efficiency of our method.