Citation

Tan, Chue Poh and Koay, Jun Yi and Lim, Ka Sing and Ewe, Hong Tat and Chuah, Hean Teik (2007) Classification of multi-temporal SAR images for rice crops using combined Entropy Decomposition and Support Vector Machine technique. Progress In Electromagnetics Research, 71. pp. 19-39. ISSN 1559-8985

Text Classification of multi-temporal SAR images for rice crops using combined Entropy Decomposition and Support Vector Machine technique.pdf Restricted to Repository staff only Download (944kB)

Abstract

This paper presents a combined Entropy Decomposition and Support Vector Machine (EDSVM) technique for Synthetic Aperture Radar (SAR) image classification with the application on rice monitoring. The objective of this paper is to assess the use of multi-temporal data for the supervised classification of rice planting area based on different schedules. Since adequate priori information is needed for this supervised classification, ground truth measurements of rice fields were conducted at Sungai Burung, Selangor, Malaysia for an entire season from the early vegetative stage of the plants to the ripening stage. The theoretical results of Radiative Transfer Theory based on the ground truth parameters are used to de. ne training sets of the different rice planting schedules in the feature space of Entropy Decomposition. The Support Vector Machine is then applied to the feature space to perform the image classification. The effectiveness of this algorithm is demonstrated using multi-temporal RADARSAT-1 data. The results are also used for comparison with the results based on information of training sets from the image using Maximum Likelihood technique, Entropy Decomposition technique and Support Vector Machine technique. The proposed method of EDSVM has shown to be useful in retrieving polarimetric information for each class and it gives a good separation between classes. It not only gives significant results on the classification, but also extends the application of Entropy Decomposition to cover multi-temporal data. Furthermore, the proposed method offers the ability to analyze single-polarized, multi-temporal data with the advantage of the unique features from the combined method of Entropy Decomposition and Support Vector Machine which previously only applicable to multi-polarized data. Classification based on theoretical modeling is also one of the key components in this proposed method where the results from the theoretical models can be applied as the input of the proposed method in order to de. ne the training sets.

Item Type:	Article
Subjects:	T Technology > T Technology (General) Q Science > QC Physics
Divisions:	Faculty of Engineering and Technology (FET)
Depositing User:	Ms Suzilawati Abu Samah
Date Deposited:	04 Oct 2011 02:44
Last Modified:	29 Dec 2020 17:48
URII:	http://shdl.mmu.edu.my/id/eprint/3123

Downloads

Edit (login required)