DETECTION AND IDENTIFICATION OF ENGINEERED SURFACES USING IMAGING SPECTROMETRY

Abstract

The urban environment in developing countries is transforming from rural to urban areas at a rapid pace. Urbanization brings economic and social benefits (e.g. economic prosperity and improved quality of life), but it also causes many environmental effects such as it directly impacts surface runoff [1] [2], degradation in water quality, loss of biodiversity and urban heat island effect [3] [4] [5], etc. Due to these major impacts, understanding the behavior of the urban environment and their spatio-temporal analysis, become essential for local and regional planning and environmental management. This demands development of some cost-efficient approaches to get urban sprawl information timely. As an indicator of urbanization, built-up surface mapping has turned out to be an active area of research and various techniques have been developed in the recent past [6] [7] [8] [9]. Therefore, taking advantage of imaging spectrometry, in this research, the detection and identification of engineered or built-up surfaces have been carried out under different objectives: i. The first objective deals with the creation of a spectral library of urban built-up surfaces and materials for Indian regions, and analysis of spectral signatures of these surfaces and materials. ii. In the second objective, the research has been carried out in four different subobjectives: a. In the first sub-objective, three new spectral indices i.e. New Impervious Index (NII), Road Detection Index (RDI) and New Roof Extraction Index (NREI) have been proposed for detection of built-up (Level-1), road and roof surfaces (Level-2), respectively. iv b. In second, two new spectral indices have been introduced, in which Condition Index-Road (CI-Road) is utilized for condition analysis of road surfaces while Deterioration Index-Roof (DI-Roof) is used for deterioration analysis of roof surfaces in AVIRIS-NG hyperspectral imagery. c. In third sub-objective, existing multispectral built-up indices formulated for extraction of built-up surfaces, have been used for extraction of urban built-up surfaces along with its sub-categories in hyperspectral imagery. d. Finally, based on existing literature, extraction of impervious or engineered surfaces has been carried out using index based Red-GreenBlue (RGB) and Near Infrared (NIR) band combinations in AVIRIS-NG imagery. iii. In the third objective, a new method is proposed, in which different combinations of feature bands have been created for extraction of built-up surfaces, sub-surfaces and materials in different levels (Level-1, 2 and 3) using AVIRIS-NG hyperspectral imagery. The knowledge based features identified in this study are thematic spectral indices, major principal components and fractional abundances. iv. In the fourth and final objective of the research, a performance evaluation of Sentinel-2B, Landsat-8 multispectral, and AVIRIS-NG hyperspectral imageries for extraction of road and roof surfaces using proposed spectral index based, and other conventional algorithms has been presented. The New Road Extraction Index (NREI) and New Building Extraction Index (NBEI) are developed for extraction of road and roof surfaces, respectively. Moreover, existing Spectral Angle Mapper (SAM), Spectral Information Divergence v (SID), Matched Filtering (MF), and Support Vector Machine (SVM) are utilized as angle, information, filtering, and machine learning based algorithms, respectively, for detection of both the surfaces. The entire analysis has been carried out using AVIRIS-NG and ground hyperspectral data of the Udaipur and Jodhpur, Rajasthan region of India. The results of the analysis depict that, indices based approach outperforms other conventional classification / detection algorithms for extraction and estimation of engineered / built-up surfaces, sub-surfaces, and materials in AVIRIS-NG hyperspectral imagery with less time and computational complexity.