EMI-NL satellite-borne imaging spectrometer utilizing charge-coupled device CCD275 is applied to monitor the global trace gases of the atmosphere. This spectrometer is the first application of CCD275 in China. Since its overall volume and power consumption are more significant than usual CCD detectors, CCD275 uses the thermal design of radiant coolers with heaters. The satellite-borne imaging spectrometer works in a sun-synchronous orbit. It captures original spectrum images during the day and dark background images at night. The original spectrum and dark background images must be captured under the same temperature and exposure time conditions. During the data processing, the dark background images are subtracted from the original spectrum images to obtain valid spectral images for subsequent retrieval of the trace gases. During the on-orbit operation of the radiant cooler, there is a risk of cooling efficiency decrease or complete failure. The risk leads to the temperature control failure of the CCD275, and collecting original spectrum images and dark background images at the same temperature becomes a problem. This paper proposes a method to calculate the dark background images instead of acquiring one to solve the problem of dark background collection at a specific temperature after thermal control failure. This method could calculate the dark background images at any specific temperature and exposure time by collecting two sets of dark background images with known temperatures and different exposure times. This paper discusses the working principle, the thermal design, and the dark background collection method of the satellite-borne spectrometer. The relationship formula, which the method mentioned above is based on, between pixel value, temperature, and exposure time is deduced from the analysis of dark current characteristics and the continuous frame transfer imaging mode. To verify this method, this paper built an experimental platform that included a vacuum system, cooling system, and imaging circuit. The platform can set the detectors working temperature and exposure time, obtaining dark background images. Statistic results show that the correlation coefficient between the calculated and measured images reaches 99.96% for a single point, and the pixel value deviation range is 0.06%~6.9%. For a full-frame image under 278 K, the mean error between the calculated dark background and the measured one is 0.23%, the variance error is -3.20%, the correlation coefficient is 99.94%, the pixel value deviation range is -0.70%~2.37%. The feature point matching rate reaches 97.9%. When temperature control fails, the calculated dark background can replace the measured one for spectral image processing.