City-scale vehicle tracking and traffic flow estimation using low frame-rate traffic cameras

Abstract

Vehicle flow estimation has many potential smart cities and transportation applications. Many cities have existing camera networks which broadcast image feeds; however, the resolution and frame-rate are too low for existing computer vision algorithms to accurately estimate flow. In this work, we present a computer vision and deep learning framework for vehicle tracking. We demonstrate a novel tracking pipeline which enables accurate flow estimates in a range of environments under low resolution and frame-rate constraints. We demonstrate that our system is able to track vehicles in New York City's traffic camera video feeds at 1 Hz or lower frame-rate, and produces higher traffic flow accuracy than popular open source tracking frameworks. CCS CONCEPTS • Computing methodologies → Tracking; • Software and its engineering → Real-time systems software.

A Data-driven System for City-wide Energy Footprinting and Apportionment

“…Besides the subway, the second most frequently used mode of transportation by commuters is by motor vehicle. For CityEnergy, we estimate motor vehicle commuting by analyzing real-time footage from traffic cameras, building on the work in Reference [50]. Real-time image feeds are publicly available from the New York City Department of Transportation, which provides 752 real-time traffic cameras covering major intersections [37].…”

“…Instead, specific features are extracted from each bounding box and matched in consecutive frames. We extract the same features as in Reference [50] into a feature vector, including the color histogram and the output from VGG16's conv3_3 layer. In consecutive frames, we compute correlation scores between pairs of feature vectors.…”

“…A final pairwise score is produced from a linear combination of the correlation scores. Last, the final matching is determined by solving the maximum weighted bipartite matching problem with the pairwise scores, as described in Reference [50]. This method enables tracking of vehicles across multiple frames and provides the vehicle flow across the major intersections.…”

Data-Driven Energy and Population Estimation for Real-Time City-Wide Energy Footprinting

“…Besides the subway, the second most frequently used mode of transportation by commuters is by motor vehicle. For CityEnergy, we estimate motor vehicle commuting by analyzing real-time footage from traffic cameras, building on the work in [35]. Real-time image feeds are publicly available from the New York City Department of Transportation, which provides 752 real-time traffic cameras covering major intersections.…”

You Only Look Once, But Compute Twice: Service Function Chaining for Low-Latency Object Detection in Softwarized Networks

“…The object identification and understanding within an ongoing video stream is based on the Computer Vision (CV) domain of real-time video analysis. Prominent examples for real-time object detection and analysis include Google Lens or smart city applications that perform video surveillance [3][4][5] or for connected autonomous cars, as illustrated in Figure 1. Especially for the latter, incorporating new sensor data such as from LIDAR and other on-board sensors that goes beyond image data alone is also attracting interest [6][7][8].…”