Satellite Image Automated Segmentation of Vegetation Using Spectral Analysis for Roadway Detection

The article presents an automated method of vegetation segmentation on satellite images to increase accuracy and efficiency of road detection. Modern methods of building and updating road maps often require significant time and computing resources, and are also prone to errors due to the human factor. The proposed approach combines spectral analysis using the normalized difference vegetation index (NDVI) and structural analysis using a modified cascade Viola-Jones classifier, which makes it possible to effective roadseparation from vegetation and other objects in complex urban scenes. Experimental studies conducted on high-resolution satellite images of WorldView-2 have shown that the optimal threshold of 0.3 ensures maximum suppression of vegetation while maintaining the integrity of the road network. The use of binary vegetation optimization reduces the area of analysis by an average of 65.8%, which significantly reduces the computational load and increases the speed of data processing. The proposed method demonstrates comparable accuracy with modern algorithms at a lower cost of resources and time. The research results allow us to recommend this approach for integration into monitoring systems of road infrastructure, transport management and urban planning, ensuring increased efficiency and scalability of satellite data analysis.