Spectral – spatial urban target detection for hyperspectral remote sensing data using artificial neural network

Abstract

Hyperspectral remote sensing is opening new gateways for a multitude of applications with an added advantage of high spectral and spatial resolution. Target detection of urban objects has gained prominence during the past decade for maintaining a pace with increasing urbanization. This paper aims to identify roads and roofs as urban targets using a hybrid approach of the spectral and spatial aspect of hyperspectral data. The work highlights a brief taxonomy of morphological operators namely, Dilation, Erosion, Opening and Closing with fused spectral signatures of urban targets considered. Artificial neural network (ANN) has been used as a machine learning measure due to its high prediction capability and its effectiveness over conventional target detection approaches.