本論文欲開發一有效辨識各環境變化的特徵抽取方法，解決目前特徵抽取皆無法同 時對各式環境變化進行有效辨識的問題。由先前實驗得知，各特徵抽取能解決之環境變 化不盡相同，所以本文提出結合多種特徵抽取─多重特徵─使特徵抽取不需倚靠大量的 訓練資料就能解決環境變化問題，提升材質辨識準確度，並對其解決環境變化之能力進 行驗證。根據實驗結果，可以得知：
1. 多重特徵須結合特性互補之特徵抽取才有意義，若結合相似之特徵抽取則無法進行 改善。
2. 特性互補之多重特徵能增加環境變化的強健性，並能夠解決更多環境變化的問題。
3. 特性互補之多重特徵能有效提升材質辨識的準確度。
4. 當具有相同材質辨識準確度的時候，特性互補之多重特徵能減少訓練資料量的使 用。
5. 特性互補之多重特徵能有效降低每次實驗結果的變異度。
由以上得知，特性互補之多重特徵能改善單一特徵抽取方式的缺點並且提升其優點，
進而提升材質辨識的準確度，減少使用的訓練資料量，並且能降低實驗結果之間的變異 度。然而目前仍有不足的地方：
1. 由於實驗中─4.3 節─選擇的三個特徵抽取太過相似，造成三個特徵結合的結果不 盡理想。未來可再進行具有互補性的三個特徵抽取結合的多重特徵，找出更有效提 升材質辨識準確度的方式。
2. 每次實驗需先測試各種權重結果，才能得知最佳權重值，造成計算時間過長。若能 找到最佳求得權重的方式，則可對實驗整體時間有效減少。
41
3. 由第五章可以得知，多重特徵能降低訓練資料量，然而當訓練資料數目過低，實驗 間的變異度也會跟著變大。若能找出較好的訓練資料選取方式，則可維持少量訓練 資料並降低變異度。
4. 由於目前實驗限定於材質辨識的特徵抽取，因此利用範圍有所限制。若能增加色彩 或是形態的特徵抽取，未來利用範圍會更為廣泛。
42
參考文獻
1. Soh, L.-K. and C. Tsatsoulis, Texture analysis of SAR sea ice imagery using gray level co-occurrence matrices. Geoscience and Remote Sensing, IEEE Transactions on, 1999.
37(2): p. 780-795.
2. Abdelrahman, A. and O. Hamid, Breast Ultrasound Images Enhancement Using Gray Level Co-Occurrence Matrices Quantizing Technique. International Journal of
Information Science, 2012. 2(5): p. 60-64.
3. Shan, C., Learning local binary patterns for gender classification on real-world face images. Pattern Recognition Letters, 2012. 33(4): p. 431-437.
4. Haralick, R.M., Statistical and structural approaches to texture. Proceedings of the IEEE, 1979. 67(5): p. 786-804.
5. Sklansky, J., Image segmentation and feature extraction. Systems, Man and Cybernetics, IEEE Transactions on, 1978. 8(4): p. 237-247.
6. Hsu, C.-W., C.-C. Chang, and C.-J. Lin, A practical guide to support vector classification. 2003.
7. Devroye, L., A probabilistic theory of pattern recognition. Vol. 31. 1996: springer.
8. Ojala, T., M. Pietikainen, and T. Maenpaa, Multiresolution gray-scale and rotation invariant texture classification with local binary patterns. Pattern Analysis and Machine Intelligence, IEEE Transactions on, 2002. 24(7): p. 971-987.
9. Haralick, R.M., K. Shanmugam, and I.H. Dinstein, Textural features for image classification. Systems, Man and Cybernetics, IEEE Transactions on, 1973(6): p.
610-621.
10. Unser, M. and M. Eden, Nonlinear operators for improving texture segmentation based on features extracted by spatial filtering. Systems, Man and Cybernetics, IEEE Transactions on, 1990. 20(4): p. 804-815.
43
11. Weszka, J.S., C.R. Dyer, and A. Rosenfeld, A comparative study of texture measures for terrain classification. Systems, Man and Cybernetics, IEEE Transactions on, 1976(4): p. 269-285.
12. Porter, R. and N. Canagarajah. Robust rotation-invariant texture classification:
wavelet, Gabor filter and GMRF based schemes. in Vision, Image and Signal Processing, IEE Proceedings-. 1997. IET.
13. Voorhees, H. and T. Poggio. Detecting textons and texture boundaries in natural images. in Proceedings of the First International Conference on Computer Vision.
1987.
14. Davis, L.S., M. Clearman, and J.K. Aggarwal. A comparative texture classification study based on generalized cooccurrence matrices. in Decision and Control including the Symposium on Adaptive Processes, 1979 18th IEEE Conference on. 1979. IEEE.
15. Haralick, R.M. and L. Watson, A facet model for image data. Computer Graphics and Image Processing, 1981. 15(2): p. 113-129.
16. Moser, G., S.B. Serpico, and J.A. Benediktsson. Markov random field models for supervised land cover classification from very high resolution multispectral remote sensing images. in Advances in Radar and Remote Sensing (TyWRRS), 2012 Tyrrhenian Workshop on. 2012. IEEE.
17. Sun, X., et al. Multi-scale local pattern co-occurrence matrix for textural image classification. in Neural Networks (IJCNN), The 2012 International Joint Conference on. 2012. IEEE.
18. Leung, T. and J. Malik. Recognizing surfaces using three-dimensional textons. in Computer Vision, 1999. The Proceedings of the Seventh IEEE International Conference on. 1999. IEEE.
44
19. Varma, M. and A. Zisserman, A statistical approach to texture classification from single images. International Journal of Computer Vision, 2005. 62(1-2): p. 61-81.