Pedestrian Trajectory Learning via Information Entropy Constrained Sparse Representation in Video Surveillance
Recognizing pedestrian behavior from video surveillance systems is an essential issue in pedestrian active safety and public safety. In this study, an information entropy constrained sparse reconstruction model is developed for pedestrian behavior learning, which aims at reducing the entropy of the reconstruction and to lead to superior classification results. Expectation Maximization (EM) method is applied to iteratively obtain the selection probabilities and the sparse coefficients. Trajectory similarity is introduced for initialization of the selection probabilities. In order to validate the performance and effectiveness of the proposed method, two datasets (including Subway dataset and Square dataset) are used in our experiments. Pedestrian behaviors can be classified more accurately using the proposed method. The classification accuracy is significantly improved compared with most well-known classifiers namely NB, kNN, SVM, and typical extant sparse reconstruction methods.
-
Supplemental Notes:
- This paper was sponsored by TRB committee ANF10 Standing Committee on Pedestrians.
-
Corporate Authors:
500 Fifth Street, NW
Washington, DC United States 20001 -
Authors:
- Chen, Zhijun
- Wu, Chaozhong
- Zhang, Yi-Shi
- Ran, Bin
-
Conference:
- Transportation Research Board 96th Annual Meeting
- Location: Washington DC, United States
- Date: 2017-1-8 to 2017-1-12
- Date: 2017
Language
- English
Media Info
- Media Type: Digital/other
- Features: Figures; Photos; References; Tables;
- Pagination: 15p
- Monograph Title: TRB 96th Annual Meeting Compendium of Papers
Subject/Index Terms
- TRT Terms: Behavior; Pedestrian safety; Pedestrians; Surveillance; Video cameras
- Subject Areas: Operations and Traffic Management; Pedestrians and Bicyclists; Safety and Human Factors;
Filing Info
- Accession Number: 01624588
- Record Type: Publication
- Report/Paper Numbers: 17-06175
- Files: TRIS, TRB, ATRI
- Created Date: Jan 30 2017 5:14PM