An adaptive background extraction method in traffic scenes

作者：Li, Xiying^*; Li, Guoming; Huang, Qiuxiao; Wang, Zhenbo; Yu, Zhi^*

来源：Optik, 2018, 156: 659-671.

DOI：10.1016/j.ijleo.2017.11.174

• Background extraction
• Running average method
• Adaptive selection
• Hierarchical learning factor

This paper presents an adaptive background extraction method in traffic scenes. Based on running average method, the proposed method uses the hierarchical learning factor which has different values in different levels to extract background. Traditionally, the running average method uses a fixed learning factor. It has some shortcomings: if the value of learning factor taken is too large, ghosts are prone to occur in the background learning process; if the value is too small, the speed of background extraction will be slow. In the proposed method, the value of image texture entropy is used to measure the convergence degree of background. According to the convergence degree, the learning factors are selected in an adaptive manner. Experiments prove that the proposed method can overcome the above shortcomings of traditional running average method and can faster extract the background without ghosts. The background composition curve analysis shows that this method is reasonable due to reasonable component distribution without excessive reliance on certain video frames. In addition, the proposed method is compared with some state-of-the-art methods. The comparison shows that our method has high detection accuracy and computation speed.

• 出版日期2018
• 单位中山大学