Title :
Real-Time Video Mining Based on SNGRLD-rLDA Model
Author :
Lin Tang ; Lin Liu ; Junhong Su
Author_Institution :
Phys. Dept., Kunming Inst., Kunming, China
Abstract :
In this paper we introduce a novel probabilistic topic model named rLDA-SNGRLD for motions or activities mining in complex scene. Based on the improvement of rLDA model, we developed SNGRLD algorithm which can inference in real-time with massive video stream data set and mine the video latent motion topics and motion regions online. Experiments prove that the application of this model for detecting and locating abnormal events in complex scene have a good real-time performance and effectiveness.
Keywords :
data mining; gradient methods; image motion analysis; object detection; probability; real-time systems; stochastic processes; video streaming; Riemannian Langevin dynamics; SNGRLD algorithm; SNGRLD-rLDA model; abnormal events detection; abnormal events location; complex scene; massive video stream data set; motion regions; probabilistic topic model; rLDA-SNGRLD; real-Time video mining; real-time performance; regional LDA model; stochastic gradient Langevin dynamics; video latent motion topics; Convergence; Data mining; Heuristic algorithms; Inference algorithms; Optimization; Stochastic processes; Streaming media; Langevin Dynamics; stochastic optimization; topic model; video mining;
Conference_Titel :
Intelligent Human-Machine Systems and Cybernetics (IHMSC), 2014 Sixth International Conference on
Conference_Location :
Hangzhou
Print_ISBN :
978-1-4799-4956-4
DOI :
10.1109/IHMSC.2014.141
