An Accurate and Convenient Method of Vehicle Spatiotemporal Distribution Recognition Based on Computer Vision.

The Convenient and accurate identification of the traffic load of passing vehicles is of great significance to bridge health monitoring. The existing identification approaches often require prior environment knowledge to determine the location of the vehicle load, i.e., prior information of the road, which is inconvenient in practice and therefore limits its application. Moreover, camera disturbance usually reduces the measurement accuracy in case of long-term monitoring. In this study, a novel approach to identify the spatiotemporal information of passing vehicles is proposed based on computer vision. The position relationship between the camera and the passing vehicle is established, and then the location of the passing vehicle can be calculated by setting the camera shooting point as the origin. Since the angle information of the camera is pre-determined, the identification result is robust to camera disturbance. Lab-scale test and field measurement have been conducted to validate the reliability and accuracy of the proposed method.