When a gas leaks, it spreads in space by diffusion, and a dynamically stable plume of concentration usually forms near the leak source, presenting an approximate "static" region in the infrared image. This characteristic often leads to the accuracy of the commonly used moving object detection algorithm in these regions is reduced, and it is difficult to obtain the spatial concentration distribution of gas. To solve this problem, an adaptive threshold detection algorithm of Vibe Gases based on background subtraction method is proposed in this paper, and two key phases of gas plume imaging are improved. In the foreground extraction stage, the foreground difference matrix is constructed by gas detection logic and two-dimensional frequency mapping is carried out. Then the difference distribution function is fitted by least square method to calculate the optimal threshold of foreground and background separation. In the background update stage, the signal matrix of the foreground gas is constructed and two-dimensional frequency mapping is carried out. The main signal range is extracted by high-pass filtering, and the pixels located in the gas region and within the main signal range are delayed updated. The infrared detection imaging experiment shows that the detection accuracy index of ethylene at 20 meters is 91.0% and the intersection ratio index is 89.4% when the gas leakage reaches stability. The detection accuracy index of small leakage sulfur hexafluoride at 5 meters is 81.3%, and the intersection ratio index is 80.7%. The algorithm significantly improves the imaging quality of the gas plume, enhances the adaptability of detection to different gases and scenes, and effectively extracts the gas concentration distribution.