The fusion of visible color (RGB) and near infrared (NIR) images takes multispectral advantage of colors from RGB image and details from NIR image. Unlike RGB images, NIR images are robust to atmospheric environments such as Rayleigh scattering and Mie scattering. In this paper, we propose long distance imaging through RGB and NIR image fusion, named LDINet. We achieve hidden texture recovery for long distance imaging based on the fusion of RGB and NIR images. We adopt pyramid feature selection to capture multiscale information in the fusion network. Since overexposure and underexposure cause a dynamic range allocation problem in RGB image, we use the attention map of RGB image to adjust contrast enhancement. We synthesize the input smoothed RGB images for training by smoothing their original RGB images, i.e. ground truth. During training, we feed the smoothed RGB images and the details of NIR images into the fusion network as input, while feeding the ground truth as output. Experimental results show that LDINet successfully recovers hidden textures lost in RGB images while keeping colors and outperforms state-of-the-art fusion methods in terms of visual quality and quantitative measurements.