臺灣博碩士論文加值系統

在本文中，我們提出一種使用延伸型限制性波茲曼機實現智慧型物件辨識系統的方法。傳統的類神經網路用於辨識物件是相當出色的，但是局部最小值的問題仍然待解決。因此，提出的方法是一個全域最小值的類神經網路。首先，將物件從攝影機拍攝的影像作分割。為了完整地描述大量的物件，低階的特徵如形狀、紋理和顏色都是必要的。在實際情形，因為低階的特徵含有雜訊，辨識的正確率不太精確。使用已訓練的延伸型限制性波茲曼機，處理的結果是對於物件分類的最佳近似解。最後，在已設計儲存一些高階的規則的知識庫，推論引擎對於結果輸出智慧型解釋。從實驗的結果，證明提出的方法是可行的。

In this paper, we propose an approach that implements an intelligent system for object recognition using Extended Restricted Boltzmann Machine (ERBM). It is excellent to recognize the objects by a typical neural network, but the problem of local minima remains to be solved. Hence, the proposed method is a neural network of global minima. First, objects are segmented from the image which is captured by the camera. In order to describe many kinds of objects completely, low-level features such as shape, texture, and color are essential. Because of some noises of low-level features, the accuracy is not precise in an actual condition. The processed result is the optimum approximate solution for object classification using the trained ERBM. Finally, an inference engine outputs the intelligent explanation for the result in the designed knowledge base which can store some high-level semantic rules. From the experimental results, it is proved that the proposed method is feasible.

摘 要.......................................................i
ABSTRACT...................................................ii
誌 謝.....................................................iii
TABLE OF CONTENTS..........................................iv
LIST OF TABLES.............................................vi
LIST OF FIGURES...........................................vii
Chapter 1 INTRODUCTION......................................1
1.1 Research Motivation.....................................1
1.2 Survey of Related Research..............................1
1.3 Overview of Proposed Approach...........................2
1.4 Thesis Organization.....................................3
Chapter 2 OBJECT SEGMENTATION...............................4
2.1 Grabcut Segmentation Algorithm..........................4
2.2 Morphological Processing...............................10
2.3 Normalization..........................................11
Chapter 3 FEATURE EXTRACTION...............................12
3.1 Shape Descriptors......................................12
3.2 Texture Descriptors....................................14
3.3 Color Descriptors......................................17
Chapter 4 EXTENDED RESTRICTED BOLTZMANN MACHINE............20
4.1 Restricted Boltzmann Machine...........................20
4.2 Extended Restricted Boltzmann Machine..................21
Chapter 5 KNOWLEDGE BASE...................................31
5.1 Structure of Knowledge Base............................31
5.2 User Interface.........................................32
Chapter 6 EXPERIMENTAL RESULTS.............................34
Chapter 7 CONCLUSIONS AND FURTHER RESEARCH.................40
7.1 Conclusions............................................40
7.2 Further Research.......................................40
REFERNCES..................................................41
APPENDIX...................................................43
Appendix A. ERBM...........................................44

[1] Carsten Rother and Vladimir Kolmogorov, Andrew Blake, “GrabCut — Interactive Foreground Extraction using Iterated Graph Cuts,” ACM Transactions on Graphics (TOG) - Proceedings of ACM SIGGRAPH, volume 23, issue 3, pp. 309-314, 2004.
[2] Peura, M. and Iivarinen, J., “Efficiency of simple shape descriptors,” In Proceedings of the third international workshop on visual form, pp. 443-451, 1997.
[3] Iivarinen J, Peura M, Särelä J, and Visa A, “Comparison of combined shape descriptors for irregular objects,” Proc.BMVC, BMVA Press, pp. 430-439, 1997.
[4] Robert M Haralick, K Shanmugam, and Its''hak Dinstein, “Textural Features for Image Classification,” IEEE Transactions on Systems, Man, and Cybernetics, volume 3, issue 6, pp. 610-621, 1973.
[5] S. B. Kotsiantis, “Supervised Machine Learning A Review of Classification Techniques,” Informatica, volume 31, number 3, pp. 249-268, 2007.
[6] David H. Ackley and Geoffrey E. Hinton, “A Learning Algorithm for Boltzmann Machines,” Cognitive Science, volume 9, issue 1, pp. 147–169, 1985.
[7] R. Salakhutdinov and Geoffrey E. Hinton, “Deep Boltzmann Machines,” In Proceedings of the International Conference on Artificial Intelligence and Statistics, volume 5, pp. 448–455, 2009.
[8] Hugo Larochelle and Yoshua Bengio, “Classification using Discriminative Restricted Boltzmann Machines,” Proceedings of the 25th international conference on Machine learning, pp. 536-543, 2008.
[9] Hugo Larochelle, Michael Mandel, Razvan Pascanu, and Yoshua Bengio, “Learning Algorithms for the Classification Restricted Boltzmann Machine,” The Journal of Machine Learning Research, volume 13, pp. 643-669, 2012.
[10] Pugh. S, Barnes. Dave, Pullan. Derek, and Tyler. L, “Knowledge based Science Target Identification System (KSTIS),” The 10th International Symposium on Artificial Intelligence Robotics and Automation in Space, 2010.
[11] E. Hyv‥onen, S. Saarela, and K. Viljanen, “Intelligent Image Retrieval and Browsing Using Semantic Web Techniques–A Case Study,” International SEPIA Conference at the Finnish Museum of Photography, 2003.
[12] Stefan Udristoiu and Anca Loredana Ion, “Image Annotation by Learning Rules from Regions Patterns,” Complex, Intelligent and Software Intensive Systems (CISIS), 2010 International Conference on, pp. 124-131, 2010.
[13] M. Stricker and M. Orengo, “Similarity of color images,” In Proceeding of SPIE on Storage and Retrieval for Image and Video Databases III, volume 2420, pp. 381-392, 1995.
[14] Geoffrey E. Hinton, “Training Products of Experts by Minimizing Contrastive Divergence,” Neural Computation, volume 14, no 8, pp. 1771-1800, 2002.
[15] Geoffrey E. Hinton, “A practical guide to training restricted Boltzmann machines,” in Technical report 2010-003, Machine Learning Group, University of Toronto, 2010.