In this research paper, an efficient method for automatic road extraction in rural and semi-urban areas is presented. This work seeks the GIS update starting from color images and using preexisting vectorial Information.
Since the input data only the RGB bands of a satellite or aerial color image of high resolution is required. The system includes different modules: data preprocessing; clustering based on standard deviation; smoothing and fillers to filtering the clustered image; thinning to make image thinner and finally junction and pendant detection to generate graph.
In the first module the color image is rectified and generates some training set from image. The second module uses a technique for clustering. The clustering algorithm is defined in order to obtain the segmented black and white image. The clustering algorithm is based on standard deviation followed by mean and variance, for an approximate standard deviation value image clustering gives a good result. In the third module the obtained clustered black and white image is filtered by a smoothing algorithm, This smoothing algorithm finds the longest connected black components and then a filling algorithm is used to fill very small white spaces which comes inside under connected longest back components. In the fourth module ZS thinning algorithm is used which made thinner the smoothed image is to see a clear view of connected components. Finally the fifth module make a graph by a junction detection algorithm, this algorithm does the job of filtering the noise from the graph to remove small edges that not necessary to road and then generate a graph showing the node and edges and by a pendant algorithm also finds the pendant vertices. The result is an extracted road network which is defined as a structural set of elements geometrically and topologically corrects.
Experimental results show that this method is efficient in extracting and defining road networks from high resolution satellite and aerial imagery.
Zaiul
