Crowd Disaster Avoidance System (CDAS) by Deep Learning Using eXtended Center Symmetric Local Binary Pattern (XCS-LBP) T
In order to avoid crowd disaster in public gatherings, this paper aims to develop an efficient algorithm that works well in both indoor and outdoor scenes to give early warning message automatically. It also deals with high dense crowd and sudden illumina
- PDF / 780,971 Bytes
- 12 Pages / 439.37 x 666.142 pts Page_size
- 89 Downloads / 194 Views
Abstract In order to avoid crowd disaster in public gatherings, this paper aims to develop an efficient algorithm that works well in both indoor and outdoor scenes to give early warning message automatically. It also deals with high dense crowd and sudden illumination changing environment. To address this problem, first an XCS-LBP (eXtended Center Symmetric Local Binary Pattern) features are extracted which works well under sudden illumination changes. Subsequently, these features are trained using deep Convolutional Neural Network (CNN) for crowd count. Finally, a warning message is displayed to the authority, if the people count exceeds a certain limit in order to avoid the crowd disaster in advance. Benchmark datasets such as PETS2009, UCSD and UFC_CC_50 have been used for experimentation. The performance measures such as MSE (Mean Square Error), MESA (Maximum Excess over Sub Arrays) and MAE (Mean Absolute Error) have been calculated and the proposed approach provides high accuracy. Keywords Crowd disaster People counting
⋅
Texture feature
⋅
Convolutional neural network
⋅
1 Introduction In reality, public safety needed places such as malls, stadiums, festivals and in public gatherings, crowd control and crowd management becomes paramount. One of the basic descriptions of the crowd status is crowd density. Counting its flow is an important process in crowd behavior analysis. It can also be used to measure the comfort level of the crowd for detecting potential risk in order to prevent overcrowd
C. Nagananthini (✉) ⋅ B. Yogameena Department of ECE, Thiagarajar College of Engineering, Madurai, India e-mail: [email protected] B. Yogameena e-mail: [email protected] © Springer Science+Business Media Singapore 2017 B. Raman et al. (eds.), Proceedings of International Conference on Computer Vision and Image Processing, Advances in Intelligent Systems and Computing 459, DOI 10.1007/978-981-10-2104-6_44
487
488
C. Nagananthini and B. Yogameena
disasters. Crowd size is the important descriptor to detect threats like riots, fights, mass panic and stampedes. In case of traditional CCTV (Closed Circuit Television) cameras, the task of monitoring the crowd level is tedious. It is because of the requirement of large number of human resources to monitor surveillance cameras constantly over a long period of time. Recently, on 24 September 2015, about 2,070 pilgrims were died due to the crowd overflow during Hajj pilgrimage in Mina, Mecca and on 14 July 2015, at least 27 pilgrims died due to the stampede caused during Maha Pushkaralu festival on the banks of Godavari River, Andhra Pradesh, India. The above mentioned stampedes are due to lack of crowd control in advance. Hence, automated techniques should be involved for observing crowd to avoid crowd crush by estimating crowd count and crowd density. The proposed system focuses on warning the authority in advance to avoid such deadly accidents due to crowd crush. The major challenges faced by crowd detection algorithms are presence of too many people in the
Data Loading...
