Abstract: To improve the accuracy of Boosting for human object detection, Boosting with kernel base classifiers, called K-Boosting, is proposed. The proposed method uses kernel function rather than linear function, as in conventional Boosting, for base classifiers. The use of kernel function makes a better decision function therefore the accuracy is improved. Experiments on human object detection application show that the accuracy is 16% improved comparing to that of conventional Boosting. The accuracy of the proposed method is comparable to that of Support Vector Machine but the computational time is comparable to that of conventional Boosting. This proposed method is very useful for development of a real time human object detection.
M. Rahmat Widyanto and Chastine Fatichah , 2008. Boosting with Kernel Base Classifiers for Human Object Detection. Asian Journal of Information Technology, 7: 183-190.