Detection Of People Stream Features Using Eigenvalue Maps
AbstractLarge crowd video surveillance is important with respect to safety. Especially in the case of unexpected events it is beneficial to be able to detect certain features like bottlenecks as quick as possible. A number of methods have been proposed to find such occurrences but accuracy is still lacking. Our research expands on a previously presented method in order to improve the detection rate of important features. This project focusses only on bottlenecks. Eigenvalue maps derived from Jacobian matrices resulting from opical flow analysis are used to find bottlenecks in people streams. An accuracy of \textgreater 80\% was obtained using a varied but small dataset. The results indicate that using eigenvalue maps for feature detection are feasible and more reliable compared to earlier proposed similar methods.
B. Solmaz, B. E. Moore, and M. Shah, â€œIdentifying behaviors in crowd scenes using stability analysis for dynamical systems,â€ Pattern Analysis and Machine Intelligence, IEEE Transactions on, vol. 34, no. 10, pp. 2064â€“2070, 2012.
I. Stijntjes, â€œAssessment of automated crowd behaviour analysis based on optical flow,â€ 2014.
S. Ali and M. Shah, â€œA lagrangian particle dynamics approach for crowd flow segmentation and stability analysis,â€ in Computer Vision and Pattern Recognition, 2007. CVPRâ€™07. IEEE Conference on. IEEE, 2007, pp. 1â€“6.
F. Santoro, S. Pedro, Z.-H. Tan, and T. B. Moeslund, â€œCrowd analysis by using optical flow and density based clustering,â€ in Proceedings of the European Signal Processing Conference, vol. 18, 2010, pp. 269â€“273.
B. Zhan, D. N. Monekosso, P. Remagnino, S. A. Velastin, and L.-Q. Xu, â€œCrowd analysis: a survey,â€ Machine Vision and Applications, vol. 19, no. 5-6, pp. 345â€“357, 2008.
S. A. Velastin, B. A. Boghossian, and M. A. Vicencio-Silva, â€œA motion-based image processing system for detecting potentially dangerous situations in underground railway stations,â€ Transportation Research Part C: Emerging Technologies, vol. 14, no. 2, pp. 96â€“113, 2006.
S. D. Khan, G. Vizzari, S. Bandini, and S. Basalamah, â€œDetecting dominant motion flows and people counting in high density crowds,â€ 2014.
H. Fradi and J.-L. Dugelay, â€œTowards crowd density-aware video surveillance applications,â€ Information Fusion, vol. 24, pp. 3â€“15, 2015.
M. Thida, Y. L. Yong, P. Climent-PÂ´erez, H.-l. Eng, and P. Remagnino, â€œA literature review on video analytics of crowded scenes,â€ in Intelligent Multimedia Surveillance. Springer, 2013, pp. 17â€“36.
B. K. Horn and B. G. Schunck, â€œDetermining optical flow,â€ in 1981 Technical symposium east. International Society for Optics and Photonics, 1981, pp. 319â€“331.
â€œBasic evaluation measures for classifier performance,â€ http://webdocs.cs.ualberta.ca/eisner/measures.html, accessed: 30-08-2015.
Authors retain copyright and grant the journal right of first publication.