The availability of traffic flow data on urban roads, as presently collected, is limited due to the sparse number of sensors over the spatially extensive network. We describe a prototype implementation of a sensor and data analysis system that could be deployed on transit or municipal vehicles to collect traffic flow data. It consists of a data collection vehicle with positioning sensors (GPS, IMU, and OEM vehicle state sensors) as well as multiple ranging sensors to monitor the ambient traffic (vertical LiDAR). Our premise is that some data on a given link is better than the current situation where most links go unobserved. The system is capable of detecting vehicles over multiple lanes, providing velocity and shape-based vehicle classification for each vehicle. We also present the results of several validation experiments, comparing the detection system against concurrent ground truth data.