Detection and localization of crowd behavior using a novel tracklet-based model

Rabiee, Hamidreza; Mousavi, Hossein; Nabi, Moin; Ravanbakhsh, Mahdyar

doi:10.1007/s13042-017-0682-8

Cited by 30 publications

(14 citation statements)

References 40 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Generally, the research on crowd behavior has focused on two major aspects: crowd behavior detection techniques and the analysis of crowd behavior in different situations. Many researchers have suggested using different techniques to detect crowd behavior, such as using video cameras[7,21,27], a tracklet-based model[28], a hybrid tracking model and a GSLM-based neural network[29], the particle entropy[30], and energy models[31], an Air Patrol Robot[32], an entropic path-integral model[33,34], and so on. In addition, a few researchers have analyzed the crowd behavior under different circumstances, such as during an earthquake evacuation[35], in a t-shaped channel[36], in a football stadium[37] and during high-stress evacuations[38].…”

Section: Introductionmentioning

confidence: 99%

Occurrence mechanism and coping paths of accidents of highly aggregated tourist crowds based on system dynamics

2019

View full text Add to dashboard Cite

The safety of highly aggregated tourist crowds is a challenging and important issue. This paper not only provided a comprehensive analysis of the accidents of highly aggregated tourist crowds but also determined the occurrence mechanism and coping paths. Based on the analysis of multiple cases, we found that the variable status of highly aggregated tourist crowds was the result of the interaction of three main elements: multisource pressure, state mutations and management responses. A series of factors interact and result in accidents, and the lack of a management response or a low-quality management response is the root cause of such accidents. A high-quality management response is a basic safety precaution for highly aggregated tourist crowds. Therefore, forming a virtuous circle of multisource pressure, state mutations and management responses is an effective path for coping with accidents.

show abstract

Section: Introductionmentioning

confidence: 99%

Occurrence mechanism and coping paths of accidents of highly aggregated tourist crowds based on system dynamics

2019

View full text Add to dashboard Cite

show abstract

“…They also used only the violence in movies dataset and obtained 96.9% accuracy. Moreover, in [42], two descriptors were used to detect and localize the abnormal behaviors; they used a simplified histogram of oriented tracklets (sHOT) combined with a dense optical flow to recognize abnormal behavior at the final result and obtained an accuracy of 82.2% for the violent crowd dataset. In [14], the authors used ViF and then classified the final prediction using SVM, where they used five-fold cross-validation for testing and obtained 82.90% accuracy for the hockey fight dataset and 81.3% for the violent crowd dataset.…”

Section: Resultsmentioning

confidence: 99%

Violence Detection Using Spatiotemporal Features with 3D Convolutional Neural Network

Ullah

Muhammad

et al. 2019

Sensors

142

View full text Add to dashboard Cite

The worldwide utilization of surveillance cameras in smart cities has enabled researchers to analyze a gigantic volume of data to ensure automatic monitoring. An enhanced security system in smart cities, schools, hospitals, and other surveillance domains is mandatory for the detection of violent or abnormal activities to avoid any casualties which could cause social, economic, and ecological damages. Automatic detection of violence for quick actions is very significant and can efficiently assist the concerned departments. In this paper, we propose a triple-staged end-to-end deep learning violence detection framework. First, persons are detected in the surveillance video stream using a light-weight convolutional neural network (CNN) model to reduce and overcome the voluminous processing of useless frames. Second, a sequence of 16 frames with detected persons is passed to 3D CNN, where the spatiotemporal features of these sequences are extracted and fed to the Softmax classifier. Furthermore, we optimized the 3D CNN model using an open visual inference and neural networks optimization toolkit developed by Intel, which converts the trained model into intermediate representation and adjusts it for optimal execution at the end platform for the final prediction of violent activity. After detection of a violent activity, an alert is transmitted to the nearest police station or security department to take prompt preventive actions. We found that our proposed method outperforms the existing state-of-the-art methods for different benchmark datasets.

show abstract

“…The number of neighbors was set as k = 5 and λ = 500 in Eq. (22). For the initial clustering we used Elkan's kmeans clustering algorithm from the VLFeat toolbox [33], which was faster than the standard Lloyd's k-means.…”

Section: Methodsmentioning

confidence: 99%

“…Sharma and Guho [28] proposed a two-step trajectory clustering approach (TCA) to segmenting crowd flow patterns; a trajectory extraction step to detect and track blocks or regions in the video, followed by a clustering step that utilized the shape, the location and the density of the trajectory in the neighborhood. Rabiee et al [22] detected abnormal behaviors from crowd scenes using a spatio-temporal tracklet based descriptor extracted from 3D patches. The tracklets were extracted by tracking randomly selected points in video frames within a short period of time.…”

Section: Related Workmentioning

confidence: 99%

Graph-based topic models for trajectory clustering in crowd videos

Ghamdi

Gotoh

2020

Machine Vision and Applications

View full text Add to dashboard Cite

Probabilistic topic modelings, such as latent Dirichlet allocation (LDA) and correlated topic models (CTM), have recently emerged as powerful statistical tools for processing video content. They share an important property, i.e., using a common set of topics to model all data. However such property can be too restrictive for modeling complex visual data such as crowd scenes where multiple fields of heterogeneous data jointly provide rich information about objects and events. This paper proposes graphbased extensions of LDA and CTM, referred to as GLDA and GCTM, to learn and analyze motion patterns by trajectory clustering in a highly cluttered and crowded environment. Unlike previous works that relied on a scene prior, we apply a spatio-temporal graph (STG) to uncover the spatial and temporal coherence between the trajectories of crowd motion during the learning process. The presented models advance the conventional approaches by integrating a manifold-based clustering as initialization and iterative statistical inference as optimization. The output of GLDA and GCTM are mid-level features that represent the motion patterns used later to generate trajectory clusters. Experiments on three different datasets show the effectiveness of the approaches in trajectory clustering and crowd motion modeling.

show abstract

Detection and localization of crowd behavior using a novel tracklet-based model

Cited by 30 publications

References 40 publications

Occurrence mechanism and coping paths of accidents of highly aggregated tourist crowds based on system dynamics

Occurrence mechanism and coping paths of accidents of highly aggregated tourist crowds based on system dynamics

Violence Detection Using Spatiotemporal Features with 3D Convolutional Neural Network

Graph-based topic models for trajectory clustering in crowd videos

Contact Info

Product

Resources

About