|M.Sc Student||Beit-Yaakov Yochai|
|Subject||Integrating Multi-Spectral Images and Aerial Photographs|
for Urban Change Detection
|Department||Department of Civil and Environmental Engineering||Supervisors||Mr. Amnon Krupnik|
|Professor Emeritus Yerach Doytsher|
Automatic detection of cartographic objects has been the subject of numerous research studies in the past two decades. The demand for such sophisticated algorithms has constantly increased, particularly in the context of GIS updating and revision.
An automated mapping process consists of two main stages: detection and extraction. Most research studies in the area have dealt with semi-automated mapping, which focuses on the extraction stage. The detection of objects is done by a human operator, who pinpoints the place where the extraction process should be applied. Fully automated mapping is supposed to integrate the detection and extraction processes into a solid autonomous process.
This research study deals with the automation of the detection phase. The study proposes a method for a fast, coarse building detection based on image classification and other advanced image processing tools, applied to a multi-spectral, 4-channel imagery, and a more detailed aerial photograph.
The result of the proposed method is a set of pointers to building hypotheses on the multi spectral imagery. In most cases, these hypotheses are correct. However, in order to reduce the error rate, a verification stage is performed by involving the information from a detailed panchromatic aerial image. While in most cases the automated procedure is capable of correctly detecting the buildings, under certain circumstances it fails to do so. Therefore, in cases of uncertainty, manual verification is used.
The verified building hypotheses can constitute
an input for existing automatic extraction processes. Full automation of the building
mapping process will enable a vast collection of up-to-date information in
frequent time intervals. This information will be utilized for change
detection and updating of existing GIS database.
The thesis describes the proposed approach in detail and presents results obtained with real images. Conclusions and future work are indicated as well.