Federal Highway Administration (FHWA), National Speed Limit Inventory

High Street and FHWA worked together to produce a National Highway System (NHS) speed limit inventory for the nation. As part of that study, High Street leveraged images taken by state DOT pavement profiler vans and trained object detection models to recognize the presence of speed limit signs in those images. The project team was able to pass millions of images through the object detection models, identify the presence and location of speed limit signs, and then extract the text of the speed limit sign using Optical Character Recognition (OCR). Ultimately, the data derived from the object detection portion of the study was combined with data derived from other machine-learning models based on roadway characteristics and observed user speeds on HPMS sample sections to create a nationwide spatial speed limit inventory.