Random vector functional-link net based pedestrian detection using multi-feature combination

Pedestrian detection is one of the key technologies in the field of computer vision. To improve the accuracy and efficiency of the recognition system, a variety of feature extraction and classification methods has been utilized on this challenging task. This paper proposes a novel multi-feature extraction and selection method to represent and distinguish different categories of samples. In addition, combined with the multi-feature combination, random vector functional-link net (RVFL) has been used to recognize these pedestrians from backgrounds. Experimental results show that multi-feature combination outperforms other widely used image features. Moreover, the performance of RVFL algorithm with multi-feature combination is even better than other state-of-the-art classification algorithms, such as SVM or AdaBoost based classifier.