A review on classifying abnormal behavior in crowd scene
群体场景下的异常行为探测方法综述
Journal of Visual Communication and Image Representation, Volume 58, January 2019, Pages 285-303
A. A. Afiq, M. A. Zakariya, M. N. Saad, A. A. Nurfarzana, M. Faizari
摘要：Crowd behavior analysis has become one of the new areas of interest in the computer vision community due to the increasing demands from surveillance and security industries. It is important to meticulously understand crowd behavior to prevent any disaster and unwanted incidents such as thief, stampede and riots. For this purpose, crowd features such as density, motion and trajectory are analyzed to detect any abnormality in the crowd. Thus, this review is aimed to provide insight on several detection methods including Gaussian Mixture Model (GMM), Hidden Markov Model (HMM), Optical Flow method and Spatio-Temporal Technique (STT). Providing the latest development, the review presented the studies that are published in journals and conferences over the past 5 years.
Chapter 14: Crowd behavior analysis from fixed and moving cameras
固定和移动摄像头视频的群体行为分析
Multimodal Behavior Analysis in the Wild, 2019, Pages 289-322
Pierre Bour, Emile Cribelier, Vasileios Argyriou
摘要：Current crowd behavior analyses based on ‘live’ visual or audiovisual camera streams require numerous personnel and is error prone due to human mistakes. Hence, automatic crowd behavior detection is urgently needed, in particular for crowded public spaces that need 24/7 monitoring. However, the complex nature of crowd behaviors and real-life constraints debase the advantages in sophisticated computer vision and analytics algorithms when they are applied in crowed monitoring systems. It is evident that crowd behavior analysis and recognition are critical to many important applications including surveillance, robotics, information retrieval, psychology, entertainment (movies, CGI, games, etc.) and market research. Automated crowd behavior analysis has been a topic of great interest in computer vision and cognitive sciences. Recently, with the growth of crowd phenomena in the real world, this area of research has attracted much attention. Therefore, over the past few years the number of works on crowd behavior analysis increased covering macroscopic (holistic) and microscopic (object-based) approaches, including techniques based on motion patterns, tracking, activity analysis and modeling, anomaly detection, and density estimation. In this chapter state-of-the-art techniques on this topic are outlined. An overview of the related challenging tasks is provided and existing and popular solutions as well as evaluation metrics and datasets are discussed.
On an algorithm for human action recognition
人类行为识别算法
Expert Systems with Applications, Volume 115, January 2019, Pages 524-534
Suraj Prakash Sahoo, Samit Ari
摘要：Human action recognition which needs video processing in real time, requires large memory size and execution time. This work proposes a local maxima of difference image (LMDI) based interest point detection technique, random projection tree with overlapping split and modified voting score for human action recognition. In LMDI based interest point detection method, difference images are obtained using consecutive frame differencing technique and next, 3D peak detection is applied on the bunch of calculated difference images. Histogram of oriented gradients and histogram of optical flow as local features are extracted by defining a block of size 16 × 16 around each of the interest point. These local features are then indexed by random projection trees. Overlapping split is used during tree structuring to reduce failure probability. Hough voting technique is applied on testing video to compute highest similarity matching score with individual training classes. In addition to Hough voting score, the number of matched interest points of a single query video with each training class, is considered for recognition. The proposed method is evaluated on segmented UT-interaction dataset, J-HMDB dataset and UCF101 dataset. The experimental results indicate that the proposed technique provides better performance compared to earlier reported techniques.
