ABSTRACT
This paper explores the idea of using a camera and image analysis to extract traffic data in congested situations. The camera system measures vehicle positions and hence car/truck mix and inter-vehicle gaps, all key data for long-span bridge loading. It also measures vehicle lengths and approximates weights from the lengths, an approach shown to give good accuracy in statistical studies of this type. A high-resolution camera, installed on FRB in 2017, captured image data at 1 second intervals over a five-month period. Standard image processing approaches are applied to extract the lengths of the vehicles from the images. Calibration factors are used to convert the lengths in the images from pixels to metres and to correct for perspective. Free-flowing traffic captured by the camera is compared with the same traffic recorded by the on-site WIM system and confirms the accuracy of the calibration process. A year of data from the WIM system is used to establish correlations between vehicle length and weight and hence to provide estimates of weights for vehicles identified in the images. The approach is used to gather bridge traffic load information including car/truck mix and inter-vehicle gaps. Bridge load effect data are fitted to Extreme Value statistical distributions to calculate characteristic maximum values. Results are shown to be sensitive to the frequency of traffic jams on the bridge.
