Feature Extraction Strategies based on Mathematical Morphology for the Analysis of Remotely Sensed Imagery

Abstract

The thesis evolves on the development of novel feature extraction methods for the analysis of remotely sensed images which are enabled to enhance the robustness and the generalization properties of the feature extraction system. Recent developments in optical data sensors mounted on-board of both space-borne and airborne earth observation platforms have led to increasing volume, acquisition speed and a variety of sensed images. Therefore the feature extraction from remotely sensed imageries is a major concern and challenge for the photogrammetry, remote sensing, and GIS communities. The extensive survey of literatures expose the shortcomings overlooked for the existing approaches utilized in the feature extraction of remote sensing images. The automated extraction of features from the remotely sensed images has been an active area of research for over a decade due to its substantial role in several application areas viz. urban planning, transportation navigation, traffic management, emergency handling, etc. Although the concept of feature extraction is relatively simple, the reliability and accuracy remains a major challenge. With advanced imaging technologies, there is an augmented demand for developing new approaches which can exhaustively explore the information embedded in remote sensing images. The past studies evidenced mathematical morphological tools as best suited for the potential exploitation of the spatial information in the remote sensing imageries. Priorly, mathematical morphology was applied only for the interpretation of binary images. However, it was extended to analyze grey scale and colour images. The thesis presents different spatial feature extraction methods which are developed based on mathematical morphology for the analysis of remote sensing optical images addressing to different applications such as urban feature detection, waterbody extraction, crop field boundary extraction and shoreline extraction. The morphology based feature extraction algorithms developed are effective and contribute to the interpretation of high resolution remotely sensed images.This automatic, scalable, and parallel processing methods can be used to analyze colossal remote sensing data within the selected classification schemes of remote sensing image system. The proposed methodologies contribute to the operational use of remote sensing datasets in manyii practical applications related to monitoring and management of environmental resources. In this thesis, a novel approach is presented for extracting shoreline from remotely sensed images. Shoreline extraction is inevitable for several studies such as coastal zone management, coastline erosion monitoring, GIS database updating, watershed definition, flood plain mapping and the evaluation of water resources. Multiple techniques are proposed for the extraction of different types of waterbodies such as lakes, rivers and glacier lakes. MM techniques have been exploited for the extraction of crop field boundaries from multiple satellite imageries. UAV driven images are beneficial as they facilitate a comprehensive description of the scenes, and concurrently require pertinent image processing techniques to exploit the geometrical information from the image datasets. This study introduces two innovative feature extraction methods for UAV and satellite images The novel feature extraction techniques proposed in the thesis have been investigated and experimented in different datasets to test their degree of performance. The experimental investigation performed with the developed techniques for analysis of remotely sensed images are noted for its improved accuracy when compared against other state of the art methods.