Machine Learning and Intelligent Communications. First International Conference, MLICOM 2016, Shanghai, China, August 27-28, 2016, Revised Selected Papers

Research Article

Intelligent Recognition of Traffic Video Based on Mixture LDA Model

Download
190 downloads
  • @INPROCEEDINGS{10.1007/978-3-319-52730-7_36,
        author={Xiaowei Tang and Xin-Lin Huang and Si-Yue Sun and Hang Dong and Xin Zhang and Yu Gao and Nan Liu},
        title={Intelligent Recognition of Traffic Video Based on Mixture LDA Model},
        proceedings={Machine Learning and Intelligent Communications. First International Conference, MLICOM 2016, Shanghai, China, August 27-28, 2016, Revised Selected Papers},
        proceedings_a={MLICOM},
        year={2017},
        month={2},
        keywords={Bayesian model Mixture LDA model Traffic video identification},
        doi={10.1007/978-3-319-52730-7_36}
    }
    
  • Xiaowei Tang
    Xin-Lin Huang
    Si-Yue Sun
    Hang Dong
    Xin Zhang
    Yu Gao
    Nan Liu
    Year: 2017
    Intelligent Recognition of Traffic Video Based on Mixture LDA Model
    MLICOM
    Springer
    DOI: 10.1007/978-3-319-52730-7_36
Xiaowei Tang1,*, Xin-Lin Huang1,*, Si-Yue Sun2,*, Hang Dong1,*, Xin Zhang1,*, Yu Gao1,*, Nan Liu1
  • 1: Tongji University
  • 2: Shanghai Engineering Center for Micro-satellites
*Contact email: xwtang@tongji.edu.cn, xlhuang@tongji.edu.cn, sunmissmoon@163.com, dh@tongji.edu.cn, mic_zhangxin@tongji.edu.cn, gaoyu1631643@tongji.edu.cn

Abstract

In this paper, an efficient unsupervised model is proposed to recognize simple actions and complex activities in traffic scenes which is named mixture LDA model. Under this framework, we use hierarchical Bayesian models are to describe three important components in traffic video: basic visual features, simple actions, and complex activities. This model adopts an unsupervised way to learn how to recognize traffic video. Moving pixels can be divided into different simple actions and short video clips can be divided into different complex activities in a long traffic video sequence, then we can achieve the purpose of recognizing different activities in the surveillance video.