Design And Implementation Of Vehicles Queue Length Dynamic Prediction System

With the continuous development of social economy and the acceleration of urbanization, China’s urban transportation is facing a severe test. Car ownership continues to rise, bringing convenience to residents to travel, but also bringing traffic congestion, environmental pollution, energy waste and other issues. Meanwhile, traffic jams are likely to cause accidents, which would cause more serious congestion. Accordingly, vicious cycle is formed.With the maturation of artificial intelligence and computer vision technology, video detection technology has been widely used in intelligent transportation systems. In urban traffic monitoring system, traffic information could be obtained from digital images by video and image processing techniques, so that the traffic monitoring could be achieved accurately and efficiently.In order to solve the problem of vehicles queuing video detection, with the backgrounds of computer vision, image processing and traffic wave theory, novel algorithms, which contains camera parameters calibration method in traffic scene, flexible window algorithm and linear prediction model of vehicles queue length, have been proposed in this paper. Thus the detecting and prediction of vehicle queue length was completed. Based on the theoretical study, a system has been designed and realized finally. Firstly, a camera parameters calibration method has been achieved in traffic scene. Combining traditional camera calibration and camera self-calibration method, the proposed method, which only used lane markings and lane width, has many advantages such as fewer calibration parameters, simpler process and better applicability for traffic scene. Secondly, video image processing technology, the core of which is flexible window algorithm, has been used to detect the vehicle queue length in this paper. Unlike the mobile window and fixed window, it is more consistent with the formation of the vehicle queuing. Then, linear prediction model of vehicles queue length, which has simple parameters and easier for system realization, has been formed based on traffic wave theory. Finally, design and implementation of a system for queue length dynamic prediction was achieved, thereby laying the foundation for future research in this field.In summary, with the camera parameters calibration method, flexible window algorithm and linear prediction model of vehicles queue length, this paper detects traffic flow on the road through the camera, which uses knowledge of computer vision and image processing, and then calculates and predicts the detected queue length. Finally, the queue length dynamic prediction system has been designed and realized. Experiments show that the system can satisfy real-time performance and stability requirements.