Synthetic aperture radar (SAR) radiometric calibration is critical for quantitative SAR applications. Cross-calibration offers significant advantages in terms of calibration frequency and timeliness by using calibrated satellites for the calibration of uncalibrated satellites. Nevertheless, the inconsistent incidence angles between the calibrated and uncalibrated satellites usually affect the calibration accuracy. Existing solutions based on empirical models were limited to specific ground target scenes and could result in overcorrection. To address this issue, this paper proposed an effective approach to mitigate the influence of the incidence angle difference based on a scene evaluation method. It combined historical target scene data and scene features to analyze the evaluation threshold and cosine power index that meet the conditions. The experimental results obtained from Sentinel-1 demonstrated that the proposed method exhibited superior correction performance in urban and bare land areas, as evidenced by a reduction in correction errors of 0.8 dB and 0.67 dB, respectively, in comparison to traditional methods.