Pedestrian Traffic Light Detection in Complex Scene Using Adaboost with Multi-layer Features

Journal of Engineering Research

Renjie Hu¹  xue-hua wu² 

[1] Nanjing Normal University;Southeast University

关键词: pedestrian traffic light detection;    Adaboost;    multi-layer features;   

DOI  :  

学科分类：社会科学、人文和艺术（综合）

来源: Kuwait University * Academic Publication Council

【 摘 要 】

In order to improve the accuracy of the pedestrian traffic light detection in complex scene, an image detection method using AdaBoost with multi-layer features is proposed. In the proposed method, the multi-layer features are adopted to characterize the pedestrian traffic lights, and the AdaBoost algorithm is used to extract the discriminative multi-layer features automatically. The multi-layer features consist of luminance and chrominance components, in which the luminancelayer features are to grasp the shape information, and the chrominance-layer features are to acquire color information. Based on the numerous features, hundreds of efficient weak classifiers are selected by the AdaBoost algorithm to construct a strong classifier. With the strong classifier, images are scanned in test procedure for detection of pedestrian traffic light. Testing results show that the proposed multi-layer features in the CIELAB color space greatly improve the accuracy of the pedestrian traffic light detection, and the proposed methods result in much better performance than the state-of-the-art machine learning methods.

【 授权许可】

Unknown