This paper discusses the application of the video image processing technology which is applied to the fire protection system. Using the characteristics of the image of the smoke when the fire broke out in the video sequence, the video monitoring scene was detected intelligently and real-time. In this paper three features of smoke were extracted: the growth of the area in the smoke spread, the irregular contour feature of the smoke region and the blurred background. These three dynamic characteristics is fused by a BP neural network to determine if there is smoke or not. The experimental results show that the algorithm in this article can identify smoke in video accurately, effectively and in real time.