Data fusion is the process of integrating information from multiple sources to produce specific, comprehensive, unified data about an entity. Data fusion is categorized as low level, feature level and decision level. This research is focused on both investigating and developing feature- and decision-level data fusion for automated image analysis and classification. The common procedure for solving these problems can be described as: 1) process image for region of interest’ detection, 2) extract features from the region of interest and 3) create learning model based on the feature data. Image processing techniques were performed using edge detection, a histogram threshold and a color drop algorithm to determine the region of interest. The extracted features were low-level features, including textual, color and symmetrical features. For image analysis and classification, feature- and decision-level data fusion techniques are investigated for model learning using and integrating computational intelligence and machine learning techniques. These techniques include artificial neural networks, evolutionary algorithms, particle swarm optimization, decision tree, clustering algorithms, fuzzy logic inference, and voting algorithms. This work presents both the investigation and development of data fusion techniques for the application areas of dermoscopy skin lesion discrimination, content-based image retrieval, and graphic image type classification.
展开▼
