Research on spatial image enhancement based on deep learning

In view of the problem that visible light image in space is affected by ambient light, the image signal-to-noise ratio is low, and the local shadow is caused when part of the target area is blocked, a method for image enhancement in space dark light condition is proposed, which can effectively enhance the target information in space low light condition. By building the space target model acquisition test environment, the space target sample image set under different lighting conditions was established. Through the deep network model based on ResNet built in this paper, the training and testing of the image sample set were completed, and the effective enhancement of the space target image under low lighting conditions was realized. In order to objectively evaluate the effect of the algorithm, compared the peak signal to noise ratio (PSNR) and natural statistics characteristics (NIQE) of the proposed algorithm with the preferred dark channel algorithm and the multi-scale Retinex algorithm, The results show that the indexes of the image results processed by the proposed algorithm are superior to the comparison algorithm. The research results can effectively improve the image quality degradation caused by insufficient illumination and illumination Angle constraints, provide high-quality data guarantee for subsequent image interpretation, and realize the overall improvement of the perception and recognition ability of the applied platform.