A study of on-road motorcycle recognition and motorcyclist''s helmet detection techniques

• Main Authors: Li-Tien Hsueh, 薛立典
• Other Authors: Chien-Wu Tsai
• Format: Others
• Language: zh-TW
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/03537111227417473857

碩士 === 銘傳大學 === 電腦與通訊工程學系碩士班 === 102 === In recent years, the traffic accident is so much in Taiwan. In violation of the locomotive project to motorcycle, that rider don''t wear helmet is very common. In the past, the number of violation is about 3 millions, so we can know the motorcyclist’s not wearing helmet is still very serious. As the current technology and research about image processing is also very much. Therefore, this thesis proposes an automatic method which can detect and recognize motorcycle and helmet from the road video. We use the road video to make real-time motorcycle recognition system and helmets detection system in this thesis. First we use background subtraction and image processing techniques to find the location of motorcycle, and calculate the aspect ratio and image size of the candidate motorcycle region as the features used in motorcycle recognition. In the helmet detection, we use the vertical projection to find the location of head, and use mask and grayscale image to recognize helmet. According to the experimental results, the accuracy of the proposed method in the recognition of motorcycle is 93.81%, and the accuracy of helmet detection is 95.16%. It is very suitable for the applications of road monitoring system.