Project: #339 Vehicle Trajectory and Gap Estimation for Conflict Prediction Progress Report - Reporting Period Ending: March 31, 2021 Principal Investigator: Alex Hauptmann Status: Active Start Date: July 1, 2020 End Date: June 30, 2021 Research Type: Applied Grant Type: Research Grant Program: FAST Act - Mobility National (2016 - 2022) Grant Cycle: 2020 Mobility21 UTC Progress Report (Last Updated: March 30, 2021, 6:19 p.m.) % Project Completed to Date: 40 % Grant Award Expended: 0 % Match Expended & Document: 0 USDOT Requirements Accomplishments We propose to build a system to identify and measure vehicle trajectories, speed, inter-vehicle gaps, and vehicle-pedestrian gaps using video streams from arbitrary traffic surveillance cameras. In the past period, we have followed our planned timeline and achieved solid progress. We have built an integrated system capable of tracking and counting vehicles going in different trajectories at complex intersections. It utilizes deep convolutional neural networks to extract visual features for reliable vehicle detection and tracking. The individual trajectories are then compared with predefined routes to determine the direction it’s going. This system is particularly useful for cities to adjust signal timing and road planning according to the density of traffic flow. To develop a system suitable for real deployment, efficiency is of equal importance to effectiveness. Our system is capable of real-time stream processing on a consumer-level desktop with multiple GPU cards. Alternatively, it can process videos in batches on a cloud hardware provider, e.g., Amazon Web Services (AWS), with the most cost-effective configuration and a small delay of result. We have been actively sharing our latest outcome with collaborators in the government and the private sector. In addition to the Federal Highway Administration, we are currently in contact with a company called "Quality Counts" about delivering our vehicle counting system at intersections. "Common Caches" is another company that we’re starting to work with, which focuses on detecting stop-sign runners as well as other public safety related innovations In the next reporting period, we will focus on the stop-sign runner detection task. This involves multiple stages from detection and tracking of vehicles to activity detection of vehicle stopping. It also requires some semantic understanding of the traffic situation to detect other law violations like this. Systems like these could greatly help to provide a safer traffic. Impacts The system that we have built is capable of reliably detecting and tracking each vehicle in a traffic surveillance view. Each trajectory is classified into a predefined route of interest for the user, e.g., going from street A to avenue B. This information reflects the density of the traffic flow in each direction and provides insights about signal timing and road planning. We are currently negotiating a potential technology transfer to Quality Counts, a company providing traffic statistics in U.S. cities. They are interested in using our system to count the vehicles going in each direction at an intersection in order to free up the large amount of human labor cost they have been spending on this. We are also in an initial phase of collaboration with Common Caches, a company providing analytics to U.S. law enforcement agencies. They are interested in expanding our system to do stop-sign runner detection. We have proposed a driving activity detection system built upon state-of-the-art activity detection systems. It can handle vehicle turning activities as well as starting and stopping. These activities will be the fundamental elements for detection of complex actions in the traffic system. Other Pyturbo v0.6.5: A pipeline system for efficient execution. 28k downloads since Jul. 2020. https://pypi.org/project/py-turbo/ AVI-R v1.3.9: A robust reader for AVI video files. 17k downloads since Jun. 2020. https://pypi.org/project/avi-r/ Outcomes New Partners Quality Counts LLC https://www.qualitycounts.net Common Caches https://www.ccaches.com Issues The Covid-19 pandemic has impacted our progress as it’s much harder to meet with collaborators and conduct field experiments in various traffic conditions. We are hoping to extend the current project by 6 months to allow for more complete and thorough research.