Paper Details

TRB 91st Annual Meeting (January 22-26, 2012)

Event Number:	652
Event Title:	Pedestrian Safety and Operations
Event Date:	Jan 24 2012 7:30PM- 9:30PM
Event Location:	Marriott, Salon 2
Event Description:
Event Agenda:

Vision-Based Recognition and Analysis of Pedestrian Abnormal Crossing Behavior (12-2522)

At present, study of pedestrian behaviors usually depends on data collection and analysis, theoretical modeling as well as simulation. By these means, traffic safety diagnosis and evaluation, signal timing of mixed traffic, road design and pedestrian-vehicle collision could be well guided. In this paper, we developed a novel method to automatically detect abnormal pedestrian crossing behavior based on video processing. Based on object trajectories data extracted by video tracking, motion patterns could be learnt automatically, which is an effective approach for understanding traffic scene as well as modeling and analyzing traffic behavior. Firstly, crossing pedestrian trajectories are obtained based on motion detection and tracking. Secondly, for estimating the crosswalk zone, we use origin points and destination points of pedestrian trajectories to construct the origin point set and the destination point set. These two separate point sets are fitted by 2-dimension Gaussian Mixed Model (2-D GMM) to extract the distribution of entry and exit zones. And then K-means algorithm is applied to acquire the initial parameters of GMM, and Expectation-Maximization (EM) algorithm is used to optimize the parameters. Based on these steps, the crosswalk region could be estimated automatically combined with hypothesis predefined. Moreover, pattern matching algorithm based on Bayesian classifier is presented for abnormal pedestrian crossing detection and analysis. At last, the entire approach proposed is implemented and tested at crosswalks in real-world. And the pedestrian crossing behaviors are analyzed. Two reasons leading to abnormal crossing are reduced and some helpful advices are given.

Authors
     Hu, Hongyu , Jilin University, China
     Qu, Zhaowei , Jilin University, China
     Li, Zhihui , Jilin University, China
     Jiang, Sheng , Jilin University, China
     Hu, Jinhui , Jilin University, China