Development of a fuzzy inference and Markovian system for the local registration of satellite and cartographic images

Abstract

We present in this article two complete procedures to solve some problems of the road networks. First, we must start with a process of road networks extraction based on the fuzzy clustering unsupervised approaches, and then we apply another approach for the local registration and deformation of a cartographic and a satellite road networks. For this aim, the idea is to segment first the sensed data and to recognize the basic urban classes (vegetation, roads, and other sectors). Then, starting from these classes, we extract the structures and the infrastructures interest by applying two algorithms of road network extraction (The Connectivity Weighted Hough Transform (CWHT), and the Regularised Shortest-Path Extraction (RSPE)), their different capabilities are applied for the characterization of streets with different width and shape. Finally, the proposed local registration method consists in translating the cartographic data into a graph model, and then defining Markov random fields (MRF) to fit the graph and the satellite image.