​

 

Projects

 

 

Automatic Traffic Monitoring and Management for Pedestrian and Cyclist Safety Using Deep Learning and Artificial Intelligence

 

Description

According to the US Department of Transportation and the Insurance Institute for Highway Safety (IIHS) Highway Loss Data Institute (HLDI), the number of traffic fatalities in the state of California was 3,623 in 2016, which is more than 9.2 deaths per 100,000 population (USDOT 2016). The city of Los Angeles alone has one of the highest rates of traffic death among large US cities. Currently, the city of Los Angeles has a strong understanding of vehicular travel at key intersections and corridors. Caltrans and local departments of transportation can optimize their traffic signal system to improve vehicular travel times using the Automated Traffic Surveillance and Control (ATSAC) System for vehicles and Regional Integration of Intelligent Transportation Systems (RIITS). However, understanding the movement of people, bicycles, and their interaction with vehicles is critical to avoiding traffic accidents and improving safety. Currently, there is no efficient automated system for monitoring the movement of pedestrians and bicyclists across the state of California and in major urban areas. Such a system could also provide valuable information about the traffic flow once implemented and calibrated.

With the advancement of technology, automated traffic monitoring has been gaining attention over the past decades. In particular, several methods have been proposed for pedestrian detection in the past couple of years . These methods have used different techniques including image and video processing algorithms, as well as machine learning techniques to detect human targets (pedestrian) through …

​

 

Real-Time Big Data Analytics for Traffic Monitoring and Management for Pedestrian and Cyclist Safety

 

Description

In this study, we design and develop an end-to-end system based on data analytics and deep learning methods to monitor, count, and manage traffic, particularly, pedestrians and bicyclists in real-time. The main objective of this research is to improve the safety of pedestrians and bicyclists, by applying self-sensed and intelligent systems to control and monitor the flow of pedestrians/bicyclists particularly at intersections. This paper proposes an effective end-to-end system for traffic vision, detection, and counting on real-time traffic videos. The developed system is evaluated on 12 hours of real video streams captured from actual traffic cameras in the city of Los Angeles. According to the results, the developed system can count the pedestrians with less than 2% error.