A Binocular Stereovision-based Vehicle Speed Measurement Optimization With Vehicle Multi-characteristic Detection

The number of motor vehicles has increased year by year,which has brought tremendous pressure to road traffic.The increasing pressure of road traffic makes the demand for intelligent traffic monitoring system more and more urgent,and the speed measurement of motor vehicles is a very important part in the intelligent traffic monitoring system.The original vehicle speed measurement system based on binocular stereo vision has the advantages of low cost,no limitation of the measurement site,and high measurement accuracy.However,since it is only based on license plate detection,there exists the problem that the speed of vehicles with license plate violation can’t be measured.In this paper,aiming at solving this problem,the optimization of binocular stereovision vehicle speed measurement with vehicle multi-characteristic detection is carried out.The main research content includes optimization of object detection algorithm,optimization of stereo matching algorithm,and optimization of speed measurement scheme and system.The main work includes:(1)A vehicle multi-characteristic data set is designed and constructed.With the vehicle multi-characteristic data set,seven modern object detection algorithms based on CNN in three categories are trained for vehicle multi-characteristic detection,and the YOLOv4 object detection algorithm based on FPN feature extraction structure is selected as the basis of the vehicle multi-characteristic detection algorithm in this paper.The YOLOv4 algorithm is improved by combing with the attention mechanism,in which the residual module in YOLOv4 is replaced by the ECA channel attention module.An improved ECA-YOLOv4 object detection algorithm is proposed,which improves the performance of YOLOv4 for vehicle multi-characteristic detection and reduces the parameter amount of the object detection algorithm as well.(2)On the basis of the proposed ECA-YOLOv4 object detection algorithm,the system’s stereo matching algorithm is updated accordingly.The vehicle characteristics used by the speed measurement system are optimally selected,and a multicharacteristic joint speed measurement algorithm based on license plate,logo and light is proposed.A binocular stereovision-based vehicle speed measurement system with vehicle multi-characteristic detection is designed.And the system performance is verified by experiments.The experimental results show that the speed measurement error of the proposed system meets the requirement of the Chinese national standard GB/T 21555-2007 in which the speed measurement error should be less than 6%.The proposed system can effectively improve the speed measurement accuracy of the original system and improve its robustness.