查詢結果分析
相關文獻
- 高光譜影像立方體於特徵空間之三維紋理計算
- 空載多光譜影像之紋理分析
- Geological Interpretation of Spot Imagery--A Case Study of the Laohushan Area in the North Qilian Mountains, Northwestern China
- 洋香瓜糖度檢測之研究--(1)影像紋理分析法
- 應用影像紋理分析及類神經網路辨識甘藍種苗之生長階段
- A New Method for Texture Analysis Using Morphological Gradient Texture Histograms
- A Multiresolution Lebesgue Decomposition Algorithm for Extracting Harmonic Feature of Texture Images
- 應用多分光影像系統及線性判別分析於固態培養基污染之檢測
- 抗核抗體免疫螢光顯影分析之處理方法評估
- 利用三維形態分析方法診斷肺臟腫瘤
頁籤選單縮合
題名 | 高光譜影像立方體於特徵空間之三維紋理計算=Three Dimensional Texture Computation in Feature Space for Hyperspectral Image Cubes |
---|---|
作 者 | 賴哲儇; 蔡富安; | 書刊名 | 航測及遙測學刊 |
卷期 | 14:4 2009.12[民98.12] |
頁次 | 頁303-316 |
分類號 | 440.98 |
關鍵詞 | 紋理分析; 灰階共生矩陣; 光譜分離度分析; 半變異元分析; 灰階共生張量場; Texture analysis; GLCM; Separability measures; Semi-variance analysis; GLCTF; |
語文 | 中文(Chinese) |
中文摘要 | 本研究提出灰階共生張量場概念,單次記錄高光譜影像立方體中三個像元的灰階出現頻率,並進行三階統計指標運算,以期產生更佳的三維特徵萃取。再者,本文研發新的演算法訂定三維紋理計算之視窗(核)尺寸,以光譜分離度分析,求得最能區別各類別組的波長取樣間距,做為光譜方向的視窗尺寸,並結合半變異元分析決定空間方向的視窗尺寸,增加三維紋理計算的分類精度。研究成果顯示,以分離度分析配合半變異元分析決定的視窗尺寸,可產生較佳的三維紋理計算成果。此外,灰階共生張量場在多數的統計指標能有效地提升特徵萃取與地物分類的成果。 |
英文摘要 | This study extended traditional GLCM (Gray Level Co-occurrence Matrix) texture analysis to a tensor field (Gray Level Co-occurrence Tensor Field, GLCTF) that uses three voxels from hyperspectral image cubes, and utilized third order statistics to extract subtle features from hyperspectral datasets. This paper also developed an algorithm based on spectral separability measures to identity the best sampling intervals of class pairs as the appropriate kernel sizes in the spectral dimension for GLCTF computation. Combining with spatial window sizes determined from semi-variance analysis, it is more appropriate to determine the size of a three-dimensional moving box for GLCTF computation to extract separable features from hyperspectral datasets for classification. Experimental results demonstrated in this paper indicate that coupling separability measures and semi-variance analysis can determine more appropriate kernel sizes for 3D texture computation. In addition, GLCTF in most indexes has better classification results in the test cases. |
本系統之摘要資訊系依該期刊論文摘要之資訊為主。