The hazards caused by gas leakage accident are multifaceted， such as environmental pollution， personnel and property loss， fire and explosion. Thermal infrared imaging is widely used as a qualitative detection technology that can realize large-scale and fast imaging. However， compared with general infrared image， the contrast of gas cloud infrared image is lower， the edge is more blurred， and it’s hard to detection. To solve this problem， this article proposed a leak detection method for low contrast gas infrared images based on mixed Gaussian background modeling. Firstly， in the preprocessing stage， time-domain adaptive interframe filtering algorithm was proposed to realize noise reduction and detail maintenance of infrared images. Then， based on spatial information and gradient information constraints， a spatiotemporal mixed Gaussian background model was proposed to achieve preliminary extraction of the foreground of leaked gas targets. Finally， to better remove interfering moving targets in foreground detection， an improved fast and robust fuzzy C-means clustering method was used to realize adaptive segmentation of gas regions. The experimental results show that at the leakage distance of 5 m， this detection algorithm can effectively improve accuracy， compensate for the problems of gas region voids， and reduce interference from other moving objects. The accuracy of gas leakage detection is between 92.3% and 96.3%， which has significant anti-interference and region segmentation capabilities compared to other algorithms.