Integration of lidar data in topographical feature extraction from very high-resolution aerial imagery

Abstract

Geospatial technology has been demonstrated as a reliable and efficient tool for monitoring of the land cover pattern for vast geographical areas. Although, the demand for the various thematic layers including landcover maps at finer scale has got increased for various applications such as urban studies, forestry and disaster management. In this paper, the utilization of LiDAR data for urban land cover classification of aerial imagery has been discussed. The study area has been classified into seven land-use/cover classes based on the textural, and spectral features using object-oriented classification approach. The applicability of various texture measures based on the gray level co-occurrence matrix along with the effect of varying pixel window has also been discussed. The classification results indicate that homogeneity texture image generated using 3 * 3 window size is best suitable for extraction of various topographical objects. The suitability of the various textural features has also been investigated. The LiDAR data has been found best suitable for the identification of small objects such as buildings, trees and vehicles over aerial imagery. The overall accuracy of the classification has been obtained as 87.21% with the kappa coefficient of 0.84. The outcome of the study can be effectively utilized for disaster management applications such as evacuation planning, damage assessment, and post-flood recovery effort.