Abstract: Hyperspectral technology has excellent application potentials in on-line large-scale detection of insulator contamination status. However, due to the influence of illumination angle and environmental occlusion in the field application environment, there is inevitably a shadow area on the insulator surface, resulting in a large error of hyperspectral detection, which makes it impossible to accurately evaluate the degree and distribution of contamination. Therefore, this study combines shadow recognition and segmentation technology to effectively achieve data correction in shadow areas. By collecting hyperspectral images of insulators under shadow occlusion of different shapes, a shadow recognition model is established by using U-Net semantic segmentation network. Further, multiple sets of insulator images with different shadow distributions are obtained by fixing the viewing angle and changing the lighting conditions, and the shadow data are corrected to the data under the same lighting conditions. The SVM-U Net model is used to realize the data correction of the shadow area in the hyperspectral image of insulator contamination detection. The average detection error caused by the shadow is reduced from 15.86 % to 3.21%.