In low illumination environment, the image quality of LiDAR images often decreases significantly due to insufficient illumination, background noise interference, signal attenuation and other factors, and the missing area presents a complex and changeable shape, including different shapes, sizes and positions. These characteristics are time-varying, resulting in low accuracy of the missing area of the image completion. Therefore, a missing area completion algorithm based on U-Net and GAN for low illumination LiDAR images is proposed. By using the encoder and decoder of U-Net network, the dual attention mechanism is added to the cross-layer connection between the lower sampling and the upper sampling, and the dynamic learning rate attenuation strategy is introduced to optimize the segmentation model of missing regions. According to GAN, the missing area of the image is completed, and the low-dimensional structure information of the image is restored by using the pre-completion model. The high-dimensional texture information of the missing area of the image is restored by the enhanced completion model. Experimental analysis shows that the proposed algorithm has a peak signal-to-noise ratio (PSNR) of 34.511dB and an information fidelity (VIF) of 0.974, which can obtain a satisfactory completion effect of the missing area of the low illumination lidar image.