A Novel Method for Road Detection Using High Resolution Satellite Images and Lidar Data Based on One Class Svm and Lbp Features

Now a days, fast extraction of road network is a challenging task especially in urban areas where roads are covered by height objects like trees, buildings, parking lots, vehicles etc. Imagery, especially high resolution image is main source for road detection as it contains rich texture and spectral information. This paper proposes a method based on merging of features of high resolution satellite images and their corresponding lidar data. The intensity of Lidar point cloud data can be used as an additional feature for road extraction as road surfaces have similar reflectance. Lidar data which has LAS 1.1 format has been taken corresponding to the high resolution satellite image. Local binary pattern is the feature extraction method used to extract features of the data. The merged features of these two undergo a one class SVM classification to increase the accuracy. The overall accuracy and kappa coefficient of the proposed method were 95.05% and 0.88 respectively. The results confirmed that this method has potential for detecting roads in urban areas using high resolution images and lidar data.