|Title:||Ecological mapping in Hong Kong with fine spatial resolution, IKONOS satellite images|
|Subject:||Hong Kong Polytechnic University -- Dissertations.|
Ecological mapping -- China -- Hong Kong.
Geographic information systems.
|Department:||Department of Land Surveying and Geo-Informatics|
|Pages:||x, 119 leaves : ill. (some col.), col. maps ; 30 cm.|
|Abstract:||Ecological mapping in the tropics is difficult due to the heterogeneity of the vegetation, the nature of the terrain which is often highly dissected, and general problem of determining ecological boundaries which may be indistinct, even to a field observer. There are no studies in the literature discussing the successful mapping of vegetation or habitats over large areas. In the last 20 years, two habitat surveys in the form of vegetation maps have been completed by Hong Kong government departments and private consultants, with inadequate accuracy and poor results. Since these previous projects used only medium spatial resolution sensors: Landsat and Satellite pour 1'Observation de la Terre (SPOT), it may be possible to produce more accurate ecological maps using the new generation of Very High Resolution (VHR) satellite sensor images. Traditionally, habitat mapping has used Aerial Photographic Interpretation (API). However, forty-five air photos are required to cover the study area, Shing Mun and Tai Mo Shan country parks in Hong Kong, compared with a single IKONOS scene. Additional advantages of IKONOS include spatial, spectral and temporal consistency. Therefore, if a suitable methodology for automatic habitat mapping can be developed, reduced costs and less processing time would be required. This study attempts to develop a methodology for detailed ecological mapping based on a suite of integrated image processing techniques, and with stated accuracy levels, for IKONOS images. Three different methodologies were chosen for evaluation; they are 1.Manual Aerial Photographic Interpretation, 2. Automated per-pixel Maximum Likelihood Classification (with texture measures) and 3. Automated "Multi-scale object-oriented segmentation with decision tree classification" (MOOSC). These three methodologies are evaluated by comparing them with GPS field points. During the study, a series of image processing techniques were investigated for their usefulness in ecological mapping. These include image fusion, spatial autocorrelation, vegetation indices and texture analysis. Habitats were mapped at two levels of detail. The most general level is the vegetation structure, or life form level which additionally includes the plantation of Melaleuca quinquenervia since in the winter time of the study and image, it is leafless and therefore structurally different from the other evergreen woody vegetation. The more detailed mapping level is termed "hybrid" since it includes both structural types and species. The results show that above 76 percent overall accuracy at general life form level and above 71 percent at detailed hybrid level were achieved using the automatic object-based approach (MOOSC) when the results were referenced to GPS field data. This finding was similar to that obtained from aerial photographic interpretation. Since GPS data collection was restricted to accessible roads and footpaths, the air photo mapping was used for further checking for more remote areas. This second method of accuracy testing between API and MOOSC is only relative. However, since approximately 90% accuracy was achieved at both general and detailed levels from MOOSC, when the air photo mapping was used as reference, this further confirms the similarity of the air photos and IKONOS-based methods. Considering both the absolute and relative accuracies together, the study indicates that the MOOSC can be used for rural habitat mapping, at an acceptable level of accuracy. This method is demonstrated to give far superior accuracy to the results from medium resolution satellite sensors and it is a viable alternative for replacing the traditional manual aerial photographic interpretation method for mapping large areas.|
|Rights:||All rights reserved|
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item: