The signal intensity of the earth at night is one-millionth of the reflected visible light intensity during the day. Small changes in the pixel response characteristics of the spaceborne low-light CCD imaging payload will significantly affect the imaging quality. The analysis of the on-orbit response characteristics of the CCD push-broom load shows that the abnormal response results in multiple bright lines along the track with different intensities in the low-light image, which have the characteristics of time-varying quantity, random position and nonlinear response. A correction method is proposed in which the spatial domain is loosely matched and the radiation domain is strictly mapped. By calculating the relative deviation of the mean radiation value of the pixels along the track, the bright line detection threshold is determined by histogram analysis and automatic detection is realized. On this basis, for each bright line, the method of establishing reference radiation value first and then sorting mapping is adopted to achieve bright line correction. In order to verify the effect of the algorithm, the low-light observation data of five typical uniform scenes including sea surface, desert, lake ice, fog and glacier, are selected for testing. The test results show that after correction, bright lines in the images disappear. The overall non-uniformity is improved by 44%, the non-uniformity of strong bright lines is relatively improved by 60%, and the signal-to-noise ratio of the typical dark background image is improved from 2 to 4.2. The method has the characteristics of real-time detection and correction pixel by pixel and is suitable for the operational radiometric correction of push-broom CCD on optical remote sensing satellites without on-board calibration devices.