A Vision-Based Vehicle Speed Warning System
Shih-Chieh Huang, Chien-Chuan Lin, Ming‐Shi Wang · 2012
In this paper, a vision based vehicle speed warning system was proposed which has been implemented on the PAC Duo embedded platform. The system consisted of a vehicle digital video recorder which used to capture the image in front of the vehicle, a hidden Markov model module to predict the next instant vehicle speed from the previous vehicle speeds, and a warning system to give the driver a warning signal when the predicted vehicle speed is larger than a predefined threshold value. The difficulty of the system is how to get the vehicle speed under the condition of without real vehicle speed provided. In this system, the vehicle speed provided for the hidden Markov model to predict the next instant vehicle speed is estimated from the images captured via the digital video recorder. Two methods have been evaluated for estimating the vehicle speed. One is using feature matching method between the two contiguous images; the other is via the pre-calculating distance matching of each pixel of the image. The testing speed range is from 0 to 110 km/hr and the processing frame rate is 5 fps. The accuracy of the speed estimation is the feature matching method: ±8%; and the other method: ±12%.