Automatic geospatial objects classification from satellite images

Abstract

LiDAR data has several advantages for classification of objects from satellite images. LiDAR data acquisition occurs in 24 h which contains height information of the objects. The morphological are used for extracting image features. As urban object detection is more difficult for shadow, bushes shrubs mixed with huts. This method gives an automatic approach for classification of the object from satellite images. It also presents an automatic approach for extraction of roads, vegetation with higher indexed and lowers indexed from the point clouds of LiDAR data. In the first step point Clouds from LiDAR data are preprocessed and then digital elevation model (DEM) are generated from that particular location. Then we have Created AOI using the normalized difference between DEM and DTM. Finally, the pixels of different objects are classified using spatial model. The experimental results are very promising. To identify terrain and non-terrain points from the raw LiDAR data an automated filtering algorithm is developed with the classification.