Abstract: Aiming at the problem of low efficiency of manual survey of changes in high consequence areas of natural gas pipelines, a change detection algorithm of multi-scale fusion in high consequence areas was proposed. Firstly, the high consequence area vectors along the natural gas pipeline were used to crop the images of two phases. Then, the images of the high consequence areas were input into the change detection algorithm to predict the feature information on building change of the entire image. Finally, the change pattern vector was generated according to the prediction result of the entire image. The multi-scale fusion module designed in the network integrated the highlevel and low-level feature information. In addition, the increased attention mechanism module was used to extract the distinguishing feature information from the channels and spatial ranges. Specifically, experiment was conducted with the 2019 and 2020 remote sensing images of Zhongwu Natural Gas Pipeline in Wuhan, with the F1 value and the recall rate reaching 0.849 and 0.838, respectively. The experimental results show that the calculated result by the change detection algorithm of multi-scale fusion meets the practical application requirements of change detection in high consequence areas.