Taxi Detection Based on Vehicle Painting Features for Urban Traffic Scenes

With rapid development of intelligent transportation systems (ITS), video-based road monitoring systems become increasingly popular. Most video processing systems concerns the detection and monitoring of all road vehicles. Few research refers to non-social vehicle detection specially, although it is critical for urban traffic management. In this paper, we present a video-based taxi detection system for urban transportation services. The taxi has obvious painting features. Firstly, vehicles are detected by localizing their license plates. The license plate is localized using its distinctive color and gradient features. Then, taxi features are extracted including the special color feature and the dome light with "TAXI" words. At last, taxi detection is performed by combining the detected licence plate and taxi features. Experiments in practical urban scenarios are carried out to evaluate the performance of our method.