| Author: | Li, Kin-wai |
| Title: | Automatic classification of tropical cyclone intensity from digital infrared satellite imagery |
| Degree: | M.Sc. |
| Year: | 1999 |
| Subject: | Meteorological satellites Satellite meteorology Cyclones -- Tropics Typhoons -- China -- Hong Kong Hong Kong Polytechnic University -- Dissertations |
| Department: | Multi-disciplinary Studies Department of Computing |
| Pages: | 80 leaves : ill. ; 31 cm |
| Language: | English |
| Abstract: | In order to solve the problems of subjectivity and experience requirement that cause by human interpretation in determining the tropical cyclone intensity from satellite imagery, this research is intended to develop an intelligent based automatic classifier to provide a more objective result in determining the tropical cyclone intensity and, hopefully, can improve the accuracy and efficiency of the Tropical Cyclone Warning issued by the Hong Kong Observatory. This research consists of three major components: (i) segmentation of tropical cyclone cloud system from infrared satellite imagery; (ii) identification and automatic extraction of features for determination of tropical cyclone intensity; and (iii) classification of tropical cyclone based on the extracted features from satellite imagery. 184 cloud systems are segmented through the techniques of Hierarchical Threshold Segmentation (HTS) and single grey shade thresholding segmentation. Because some tropical cyclone cloud systems are too poor in organization to be segmented, the successful rate of segmentation is only 80%. A total of 16 features are identified to depict the intensity of a tropical cyclone. The feature set consists of 14 physical features and 2 textural features. Two image processing techniques are used in feature identification and extraction, which are Edge Detection Technique and Eye Detection Technique. Two classification techniques are adopted in the implementation of automatic classification system, which are Neural Network Approach and Statistical Technique for Extracting Classificatory Knowledge. The overall performance is 61% for the neural network classifier and 64.8% for the statistical technique classifier. A test of human expert classification is conducted to verify the performance of two automatic classifiers, and the overall performance is only 53.4%. The moderate performance of two automatic classifiers may be due to the inherent limitation of identified features in representation of tropical cyclone cloud systems. However, a low performance of human expert classification implies that individual infrared satellite imagery has an insufficiency in nature to provide information to classify tropical cyclone intensity with a promising result. On the other hand, the performance characteristic on individual class for human expert classification is similar to that for two automatic classifiers. This phenomenon reveals that, to a certain extent, the identified features for classification of tropical cyclone intensity are quite effective in simulation of human perceptiveness or sensation. The accuracy of the operational technique of objective satellite analysis used by the Hong Kong Observatory is around 70%. With some improvements on feature identification in future, it is highly possible for the automatic classifiers suggested in this study to be adopted by the Hong Kong Observatory as an operational tropical cyclone classification techniques. |
| Rights: | All rights reserved |
| Access: | restricted access |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| b14605260.pdf | For All Users (off-campus access for PolyU Staff & Students only) | 3.01 MB | Adobe PDF | View/Open |
Copyright Undertaking
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:
https://theses.lib.polyu.edu.hk/handle/200/1041