Abstract: This paper focuses on the real-time and effectiveness issue of fire control system which is based on video in large space building, then proposes a camera self-calibration of the adaptive differential evolution algorithm which is based on improvement. It gets the exact matching points by using a fused method SURF fire image feature point matching algorithm that combined with ratio method and correlation function method. It processes the camera self-calibration, uses relatively accurate matching points to obtain basic matrix F, it gets the optimization function according to the property of absolute conic. It optimizes the function based on improved differential evolution algorithm, obtains the camera intrinsic parameters and gets the three-dimensional information of the fire source. Experimental results show that this method can get the relatively accurate three dimensional information about the fire source in a short period of time, in addition, the real-time and accuracy can meet the standard of the fire control system to put out the fire effectively.

Key words: correlation function, differential evolution, camera self-calibration, video fire

摘要： 针对基于视频的大空间建筑火灾消防存在的实时性和有效性问题，提出了一种基于改进的自适应差分进化算法的摄像机自标定方法。用基于比值法与相关函数法融合的SURF火灾图像特征点匹配算法，快速得到准确的匹配点对;进行摄像机自标定，用较为准确的匹配点对求得基本矩阵F，利用绝对二次曲线的性质，得到优化函数。利用基于改进的差分进化算法对其进行优化，求得摄像机内参数，得到火源的三维信息。实验结果证明，该方法短时间内，算出了较为准确的火源的空间三维信息，实时性和精确度均能够满足火灾消防的标准，有效地进行灭火。

关键词: 相关函数, 差分进化, 摄像机自标定, 视频火灾