Classification with NormalBoost: Case Study Traffic Sign Classification

NormalBoost is a new boosting algorithm which is capable of classifying amulti-dimensional binary class dataset. It adaptively combines several weakclassifiers to form a strong classifier. Unlike many boosting algorithmswhich have high computation and memory complexities, NormalBoost is capableof cl...
Ausführliche Beschreibung

NormalBoost is a new boosting algorithm which is capable of classifying amulti-dimensional binary class dataset. It adaptively combines several weakclassifiers to form a strong classifier. Unlike many boosting algorithmswhich have high computation and memory complexities, NormalBoost is capableof classification with low complexity.The purpose of this paper is to present NormalBoost as a framework whichestablishes a platform to solve classification problems. The approach wastested with a dataset which was extracted automatically from real-worldtraffic sign images. The dataset contains both images of traffic signborders and speed limit pictograms. This framework involves the computationof Haar-like features of these images and then employs the NormalBoost classifier to classify these traffic signs. The total number of images whichwere classified was 6500 binary images. A -fold validation was invoked tocheck the validity of the classification which resulted in a classificationrate of 98.4% and 98.9% being achieved for these two databases. Thisframework is distinguished by its invariance to in-plane rotation of theimages under consideration. Experiments show that the classification rateremains almost constant when traffic sign images with different angles ofrotations were tested. Ausführliche Beschreibung