Evaluation of Automatic Feature Extraction Techniques from Imagery

Abstract

Features extraction from raster images is very important of many GIS activities such as GIS updating, geo-referencing, and geospatial data integration. This process depends heavily on human labor, which makes GIS database development an expensive and time-consuming operation when performed manually. Automated feature extraction can significantly reduce the time and cost of data acquisition and update, database development and turnaround time. For the last ten years, some of the Gaza Strip municipalities and institutions have begun using geographic information system (GIS) in their projects. Therefore, it was become necessary to deal with available data from aerial images and extract features to be used in analysis, planning and decision-making. Nowadays in Palestine, all methods used in feature extraction are still traditional (manual methods). Recently, the need for using remote sensing data to accomplish the complex task of automatic extraction of features has significantly increased. Among the sensor systems currently used for mapping can be highlighted the recent launches of new orbital satellites. Extracting cartographic objects from images is a difficult task because aerial images are inherently noisy, complex, and ambiguous. This thesis reviews and provides an overview of the types of imagery being used for feature extraction. It also describes the methods used for feature extraction as well as the quantitative and qualitative accuracy assessment of these procedures. Number of optimization techniques that have been developed in stand-alone programs are studied such as ERDAS Imagine, ENVI and Barista to automate the extraction with evaluating and comparing the feature extraction workflow of these programs and feature extraction results. Finally, it is recommended to continue working on the development or improvement of existing algorithms to enhance the percentage of accuracy and speed of data, which are extracted as much as possible and the cooperation with local and global teams in this field.