a camera assembly mounted on a support structure and positioned in relation to a road surface, said camera assembly comprising a plurality of cameras configured to capture video streams of vehicles traversing said road surface;

a server assembly comprising a control hub interfaced with said camera assembly, wherein said control hub is mounted on said support structure; a processor; a memory storing machine learning modules, said machine learning modules comprising:

a multi-object tracking module configured to generate vehicle trajectories from said synchronized video streams, wherein said multi-object tracking module:

applies deep learning object detectors to localize vehicles in each video frame across multiple viewpoints;

associates detections of the same vehicle across subsequent video frames using re-identification models; and

stitches the associated detections into coherent vehicle trajectories representing paths taken by each vehicle over time using multi-object tracking algorithms;

an event correlation module configured to correlate events associated with said vehicles captured by said camera assembly, and associate said events with said vehicle trajectories;

said memory containing instructions that when read by the processor causes the system to:

track vehicles across said video streams using said multi-object tracking module to generate said vehicle trajectories;

correlate events associated with said vehicles captured by said camera assembly using said event correlation module, and associate said correlated events with corresponding vehicle trajectories;

synthesize vehicle profiles from said correlated events and associated vehicle trajectories to determine a toll, wherein said vehicle profiles comprising vehicle metadata captured by the plurality of cameras including license plate numbers, make, model, and axle count.