Abstract: The algorithm of SIFT is widely used in three-dimensional recognition. Because of its dependence on too many feature points, and intensive compute, in practice, it is difficult to ensure the real-time of the fire image. The algorithm keeps the original SIFT algorithm with good rotation, scale, lighting invariant and so on. So as to improve the speed of algorithm, it uses 4×4 window（16 vector） of key points to represent feature description. It uses the two-way match to improve the accuracy. The experimental data show that the improved algorithm has better real-time and accuracy to satisfy the need of the fire space location.

Key words: Scale Invariant Feature Transform（SIFT）, fire video image, feature matching

摘要： SIFT算法广泛用于三维目标识别，但其由于依赖过多的特征点，计算量大，在实际火灾图像定位中很难保证算法的实时性。在保留了SIFT算法良好的旋转、尺度、光照等不变特性的基础上，采用关键点的4×4窗口（16维向量）表示关键点的特征描述。同时，利用双向匹配在特征点的匹配上保证了准确率。实验数据表明，该算法在很大程度上提高了图像的匹配速度，准确率满足火灾空间定位的需要。

关键词: SIFT算法, 火灾视频图像, 特征点匹配