Abstract: A SIFT（scale invariant feature Transform） algorithm and an improved RANSAC（random sampling consistency） algorithm were proposed to solve the problem of error detection and miss detection in forest fire image recognition. The algorithm can greatly improve the matching accuracy and shorten the matching time. In this paper, SIFT algorithm was firstly used to extract feature points in the image, and then by reducing the number of points set in RANSAC, the improved RANSAC algorithm was used to process these feature points and remove the mismatched points to achieve accurate matching. The data set was obtained through burning experiment and simulated by MATLAB. By comparing the matching time and matching accuracy of different algorithms, it was proved that the improved RANSAC algorithm in this paper had the effect of improving detection efficiency for forest fire image recognition.