| 摘要 | 第6-7页 |
| Abstract | 第7页 |
| 1 INTRODUCTION | 第11-14页 |
| 1.1 Challenges and goal | 第12页 |
| 1.2 Thesis contributions | 第12页 |
| 1.3 Organization of the work | 第12-14页 |
| 2 IMAGE SEGMENTATION AND RECOGNITION’S BACKGROUND | 第14-26页 |
| 2.1 Region based image segmentation method | 第14-15页 |
| 2.2 Edge based image segmentation method | 第15页 |
| 2.3 Image segmentation method based on mathematical morphology | 第15-16页 |
| 2.4 Image segmentation method based on Fuzzy Theory | 第16页 |
| 2.5 Image segmentation method based on Neural Network (NN) | 第16-17页 |
| 2.6 Image segmentation method based on support vector machine (SVM) | 第17-18页 |
| 2.7 Image segmentation method based on graph theory | 第18-19页 |
| 2.8 Image segmentation method based on immune algorithm | 第19-20页 |
| 2.9 Image segmentation method based on Granular Computing Theory | 第20-21页 |
| 2.10 The Region Based Convolutional Neural networks | 第21-26页 |
| 3 METHODS FOR THE CLASSIFICATION AND THE SEGMENTATION | 第26-41页 |
| 3.1 Convolutional Neural Networks | 第26-35页 |
| 3.1.1 Architecture of the CNN | 第26-30页 |
| 3.1.2 The back-propagation Algorithm | 第30-35页 |
| 3.2 Data clustering methods for segmentation | 第35-39页 |
| 3.2.1 K-means clustering | 第36-37页 |
| 3.2.2 Fuzzy C-means clustering | 第37-39页 |
| 3.3 Implementation tools | 第39-40页 |
| 3.4 Summary | 第40-41页 |
| 4 EXPERIMENTATIONS AND RESULTS | 第41-53页 |
| 4.1 Training and testing Dataset preparation | 第41-43页 |
| 4.1.1 Labeling of the grayscale images dataset | 第41-42页 |
| 4.1.2 Data preparation to fed the network | 第42-43页 |
| 4.2 Presentation of the network’s architecture | 第43-47页 |
| 4.2.1 Different layers | 第43-46页 |
| 4.2.2 The general architecture of the CNN | 第46-47页 |
| 4.3 The training process | 第47页 |
| 4.3.1 Initialization of the network | 第47页 |
| 4.3.2 Batches training | 第47页 |
| 4.4 Testing process | 第47-48页 |
| 4.5 Combining the CNN and the Fuzzy C-means | 第48-49页 |
| 4.5.1 Classification with the CNN | 第48页 |
| 4.5.2 The segmentation with the Fuzzy C-means | 第48页 |
| 4.5.3 The bounding box | 第48-49页 |
| 4.6 Results of the experimentations | 第49-51页 |
| 4.6.1 Training and testing Batch of different sizes | 第49-50页 |
| 4.6.2 Applying the Fuzzy C-means for segmentation | 第50-51页 |
| 4.6.3 Final result | 第51页 |
| 4.7 Summary | 第51-53页 |
| 5 Conclusion and Future work | 第53-55页 |
| References | 第55-60页 |
| Scientific Research Published During His Master’s Degree Study | 第60-61页 |
| Acknowledgement | 第61-62页 |
