[11.16-17] Object detection by color histogram-based fuzzy classifier with support vector learning

Neurocomputing 2009
National Chung-Hsing University, Taiwan

내용
기본적으로 histogram + template matching.
Contributions:

1. The proposed non-uniform partition (in histogram)
2. The use of Self-organizing Takagi-Sugeno-type fuzzy network with support vector learning (SOTFN-SV) classifier
3.  A splitting K-means clustering algorithm is proposed (to eliminate false alarms and determine object location).

비교:

1. template matching method using color histogram. histogram intersection (HI).
2. OpenCV (Haar-like features)
3. SIFT (grey)
4. SONFIN (classifier 비교 - 이런 방법들이 등장하는 건 역시 neural networks를 다루는 저널이기 때문에)
5. Gaussian-kernel based SVM (역시 classifier)