5th International ICST Conference on Communications and Networking in China

Research Article

Moving cast shadow elimination based on luminance and texture features for traffic flow

Download301 downloads
  • @INPROCEEDINGS{10.4108/chinacom.2010.68,
        author={Liang Gao and Jianping Xing and Hui Li and Yongzhi Wang and Lina Zheng and Xiling Luo},
        title={Moving cast shadow elimination based on luminance and texture features for traffic flow},
        proceedings={5th International ICST Conference on Communications and Networking in China},
        publisher={IEEE},
        proceedings_a={CHINACOM},
        year={2011},
        month={1},
        keywords={Gaussian Mixture Model moving cast shadow detection CIE Luv color space texture analysis},
        doi={10.4108/chinacom.2010.68}
    }
    
  • Liang Gao
    Jianping Xing
    Hui Li
    Yongzhi Wang
    Lina Zheng
    Xiling Luo
    Year: 2011
    Moving cast shadow elimination based on luminance and texture features for traffic flow
    CHINACOM
    ICST
    DOI: 10.4108/chinacom.2010.68
Liang Gao1, Jianping Xing1,*, Hui Li1, Yongzhi Wang1, Lina Zheng1, Xiling Luo2,*
  • 1: School of Information Science and Engineering, Shandong University, Jinan China
  • 2: School of Electronics Information Engineering, Beihang University, Beijing, China
*Contact email: xingjp@sdu.edu.cn, luoxiling@buaa.edu.cn

Abstract

A new algorithm namely moving cast shadow elimination based on luminance and texture features (MSELT) to detect moving shadows of vehicles is investigated in this paper. Different from traditional methods only performed in color space, we combine the luminance in the CIE Luv color space and texture feature to determine shadows. The proposed algorithm based on Gaussian Mixture Model (GMM) uses the luminance weight in the CIE Luv color space to model background, do texture analysis and detect shadows. Texture analysis is performed by evaluating the gradients in the foreground with the observation that shadow regions present smooth texture characteristics. The experimental results show that this method outperforms results obtained with color space information alone, particularly in detection of vehicles which present similar luminance characteristics with shadows.