| Vehicle speed measurement has always been one of the very important tasks in intelligent traffic management.In recent years,with the rapid development of vehicle speed measurement technology,video-based vehicle speed measurement technology stands out in the research of intelligent traffic management vehicle speed measurement related technologies.At present,video-based vehicle speed measurement technology mostly relies on equipment such as road surveillance cameras,which can’t achieve vehicle monitoring and speed measurement at any location and any time due to fixed installation location,complex operation and continuous maintenance.As a necessary item in people’s lives,mobile phones have the characteristics of powerful shooting functions,small size,easy to carry,etc.,and can shoot video anytime,anywhere.Therefore,this paper designs and develops a video vehicle speed measurement system with simple structure and convenient operation for video images taken by mobile phones,which provides a more practical detection method for improving the vehicle speed detection means of traffic management,which has important research significance and high practical application value.Firstly,in order to avoid the influence of Intra-frame jitter on the speed measurement accuracy of videos shot by mobile devices such as mobile phones,a digital video stabilization processing method based on feature tracking is proposed.Based on the KLT(Kanade-Lucas-Tomasi)algorithm of image pyramid optimization to track the feature points extracted by Shi-Tomasi algorithm,this method completes global motion estimation and motion filtering by using the homological transformation model and Kalman filter algorithm based on singular value decomposition,and then uses the motion compensation method of adjacent reference frames to realize the image stabilization processing of the video.The subjective and objective evaluation of the image stabilization results shows that this method has good image stabilization effect and excellent image stabilization ability.Secondly,in order to obtain the running trajectory of the specified single target vehicle in the two-dimensional image,the normalized cross-correlation image matching algorithm based on gray-scale is used to complete the tracking of the target vehicle.In view of the shortcomings of this algorithm with large computation amount and slow matching speed,an optimal search strategy is proposed on the basis of the proximity search strategy and the hierarchical search strategy based on image pyramid.While maintaining pixel-level accuracy,the optimized algorithm has greatly improved the matching speed,which basically meets the real-time requirements of target tracking.In addition,since the detection of the actual distance in three-dimensional space according to the two-dimensional image information needs to be realized by camera calibration,a camera self-calibration method based on two vanishing points and a calibration line is proposed.In this method,the detection and screening of straight lines in the image are first completed by LSD linear detection algorithm and hierarchical clustering algorithm,and then the extraction of orthogonal vanishing point is completed by combining the RANSAC algorithm,and then the camera calibration is realized according to the characteristics of the vanishing point and a calibration line of known actual length.This method has strong flexibility and less dependence on the scene,and is suitable for cameras such as mobile phone cameras that are not easy to achieve early calibration,and the verification of examples shows that it has high detection accuracy.Finally,the construction of the video speed measurement system in this paper is completed through the MATLAB software platform,and the accuracy of the system speed measurement is verified by experiments.The experimental results show that the error of the system is maintained within the error range of 5%,which meets the accuracy requirements of vehicle speed measurement. |
PDF Full Text Request