Abstract: This paper presents a smoke foreground segmentation algorithm based on temporal-spatial analysis together in early forest fire smoke image sequences.Firstly，we analyze the instantaneous dynamical data and cumulated dynamical data.Secondly，after binarization and morphology processing，we obtain the smoke foreground contour.And then，the author can extract the smoke foreground by scanning the smoke foreground contour and filling the smoke foreground mask.Experiment results show that this method can successfully segment the smoke foreground from the early forest fire smoke image sequences with less time.

Key words: image segmentation, temporal-spatial field, instantaneous-cumulated dynamical data, smoke foreground

摘要： 针对早期森林火灾烟雾图像序列，提出了一种基于时间-空间域联合的烟雾前景分割算法。首先，对烟雾的瞬时动态数据和累积动态数据进行分析，然后，进行二值化、形态学处理，得到了烟雾前景轮廓。通过扫描烟雾前景轮廓和填充烟雾前景掩模后即可得到烟雾前景图像。实验证明，该算法兼顾了分割效果和实时性，较好地对森林火灾烟雾进行了提取。

关键词: 图像分割, 时间-空间域, 瞬时-累积动态数据, 烟雾前景