when the sky is not in the field of view, the atmospheric correction algorithm based on independent component analysis is proposed. According to the independence of the high frequency information of the atmosphere and scene light, the polarization degree of atmosphere is estimated by independent component analysis and mutual information optimization principle, and the airlight intensity at infinity is estimated with prior knowledge of the source image and atmospheric scattering physical model. Through experimental contrast analysis, the obtained atmospheric information results match the theoretical value, and the reasonability of atmospheric information estimation is verified. Meanwhile, for the influence of sensing image degradation caused by atmospheric medium, the linear polarization degree is fixed based on fuzzy rule, the airlight information is corrected, and the image quality is improved. The polarization remote sensing image reflects the terrain target characteristics more accurately after correction, and the ability of target detection and recognition is improved.