04-11-2014, 08:00 AM
Although the motorcycle is a popular and convenient means of transportation, it is difficult for researchers in intelligent transportation systems to detect and recognize, and there has been little published work on the subject. This paper describes a real-time motorcycle monitoring system to detect, recognize, and track moving motorcycles in sequences of traffic images. Because a motorcycle may overlap other vehicles in some image sequences, the system proposes an occlusive detection and segmentation method to separate and identify motorcycles. Since motorcycle riders in many countries must wear helmets, helmet detection and search methods are used to ensure that a helmet exists on the moving object identified as a motorcycle. The performance of the proposed system was evaluated using test video data collected under various weather and lighting conditions. Experimental results show that the proposed method is effective for traffic images captured any time of the day or night with an average correct motorcycle detection rate of greater than 90% under various weather conditions.