Design of multiple classifier systems based on SFLS

A novel multiple classifiers fusion approach based on SFLS (Shortest Feature Line Segment) is proposed in this paper. SFLS is a kind of simple yet effective classification method depending on the shortest feature line. The original form of SFLS´s output is just class label. To use SFLS as the member classifier in the multiple classifier system, the form of SFLS´s output is modeled using the membership function. According to the experimental results, one can conclude that the proposed multiple classifier system can effectively improve classification accuracy.