一個對於高光譜影像的模擬退火特徵齊一化波段選取方式

A Simulated Annealing Feature Uniformity Band Selection Approach for Hyperspectral Imagery

黃彥誠(Yen-Cheng Huang)

指導教授 ： 張陽郎

國立臺北科技大學/機電學院/通訊與資訊產業研發碩士專班/碩士(2008年)

摘要

高光譜遙測影像是近幾年來相當受到重視的一門技術，『高光譜』解析度之感測器已經普遍地應用於衛星遙測影像之識別、醫學影像的診斷檢查、工業產品之檢驗、飛機及其他精密機器設備之非破害性檢查等之應用，關於這方面的研究，在全球化競爭的趨勢下，已經積極的拓展開來。本篇論文提出一個新型的模擬退火特徵齊一化(SAFU)波段選取法應用在高光譜遙測影像的特徵萃取，模擬退火波段選取(SABS)法已經成功的應用在高光譜影像的特徵萃取藉著群集高度相關的高光譜波段將之形成一個基於模擬退火(SA)演算法波段模組的子集合，本文提出之SAFU是建立在SABS之上，利用舊有SABS的特性加以改良，提出SAFU來對於高光譜影像可以更有效率的進行特徵萃取。本文提出的SAFU選取不同類別抽取出不同的波段，同時地利用不同類別在高光譜影像中固有的分離性來縮減維度及更進一步地有效產生一個獨特的SAFU特徵，最後實驗列舉了一些分類正確率有不錯表現的分類器與PBF分類器進行一些正確率的比較，實驗的結果顯示SAFU方法是有效且能夠被當作目前現有的特徵萃取演算法中的其中一種。

關鍵字

模擬退火特徵齊一化 ； 模擬退火波段選取 ； 模擬退火

並列摘要

The technologys of hyperspectral remote sensing images are considerable attention in recent years. This hyperspectral resolution of the sensor has been widely used in the identification of satellite remote sensing imaging, medical imaging diagnostic tests, and testing of industrial products, aircraft and other sophisticated equipment at non-damaged inspection of the application. In this regard, the research trend of globalization of competition has been positively open up. In this paper a novel simulated annealing feature uniformity (SAFU) band selection approach is proposed for the feature extraction of the hyperspectral remote sensing images. The simulated annealing band selection (SABS) method has been successfully applied to the feature extraction of the hyperspectral images by clustering highly correlated hyperspectral bands into a smaller subset of band modules based on simulated annealing (SA) algorithm. SAFU built by SABS is an efficient feature extraction method. SABS can be improved by SAFU for Hyperspectral imageries. The proposed SAFU selects the sets of uncorrelated hyperspectral bands for each different class simultaneously while utilizing the inherent separability of different classes in hyperspectral images to reduce dimensionality and further to effectively generate a unique SAFU feature. The experimental results show that the SAFU approach is effective and can be used as an alternative to the existing feature extraction algorithms.

並列關鍵字

simulated annealing feature uniformity (SAFU) ； simulated annealing band selection (SABS) ； simulated annealing (SA)

參考文獻

[24] 林均傑， 一個對於高光譜影像的二維模擬退火波段選擇方法，碩士論文，國立台北科技大學，台北，2007年6月。

[1] J. Harsanyi and C.-I. Chang, “Hyperspectral image classification and dimensionality reduction: anorthogonal subspace projection approach,” IEEE Trans. Geosci. Remote Sensing 32, no. 4, 1994, pp. 779–785.

[2] C. Lee and D. A. Landgrebe, “Analyzing high-dimensional multispectral data,” IEEE Trans. Geosci. Remote Sensing 31, no. 4, 1993, pp. 792–800.

[3] Rudy Setiono and Huan Liu, “Neural-Network Feature Selector”, IEEE TRANSACTIONS ON NEURAL NETWORKS, VOL. 8, NO. 3, MAY 1997

[4] Yang-Lang Chang, Chin-Chuan Han, Kuo-Chin Fan, K.S. Chen, Chia-Tang Chen and Jeng-Horng Chang, "Greedy Modular Eigenspaces and Positive Boolean Function for Supervised Hyperspectral Image Classification," Optical Engineering, Vol. 42, Issue 9, pp. 2576-2587, September 2003 (SCI, EI).