Abstract: To solve the problem that existing k distribution model is not suitable for calculating the radiation characteristics of gases with drastic changes in temperature and the proportion of participating components, a multiscale multigroup wide band k-distribution model(MSMGWB) was established. The spectral absorption coefficient grouping of MSMGWB model was optimized according to the thermodynamic parameters of high temperature gas and ambient air for the calculation of remote sensing thermal image of exhaust system of marine turbine engine. The calculation results of a series of one-dimensional gas radiation transmission examples show that when the temperature of high temperature gas and low temperature air and the mole ratio of water vapor, carbon dioxide and carbon monoxide are greatly different, the prediction accuracy of the multiscale multigroup wide band k-distribution model for the 3-5 micron wave band gas radiation transmission characteristics is much higher than that of the original wide band k-distribution model and narrow band model, and it maintains good compatibility with the solid wall. Finally, the flow field, solid temperature field and close-range and long-range infrared images of exhaust system of marine turbine engine under different relative wind speeds were studied. The results show that the effects of gas gravity and carbon monoxide on infrared images under the 3-5 m band cannot be ignored.