Accurately perceiving changes in water level information is key to achieving the fine control of water and flooding; however, the existing technology cannot achieve water level recognition in complex and harsh environments, such as at night; in haze, rain, or snow; or during obscuration by floating objects or shadows. Therefore, on the basis of a deep analysis of the characteristics of water level images in complex and harsh environments, in this study, we took full advantage of a deep learning network’s ability to characterise semantic features and carried out exploratory research on water level detection in no-water-ruler scenarios based on the two technical means of target detection and semantic segmentation. The related experiments illustrate that all the methods proposed in this study can effectively adapt to complex and harsh environments. The results of this study are valuable for applications in solving the difficulties of accurate water level detection and flood disaster early warnings in poor-visibility scenarios.

The accurate identification of coal and rock at the mining face is often hindered by adverse underground imaging conditions, including poor lighting and strong reflectivity. To tackle these issues, this work introduces a recognition framework specifically designed for underground environments, leveraging joint migration and enhancement of multidimensional and full-scale visual representations. A Transformer-based architecture is employed to capture global dependencies within the image and perform reflectance component denoising. Additionally, a multi-scale luminance adjustment module is integrated to merge features across perceptual ranges, mitigating localized brightness anomalies such as overexposure. The model is structured around an encoder–decoder backbone, enhanced by a full-scale connectivity mechanism, a residual attention block with dilated convolution, Res2Block elements, and a composite loss function. These components collectively support precise pixel-level segmentation of coal–rock imagery. Experimental evaluations reveal that the proposed luminance module achieves a PSNR of 21.288 and an SSIM of 0.783, outperforming standard enhancement methods like RetinexNet and RRDNet. The segmentation framework achieves a MIoU of 97.99% and an MPA of 99.28%, surpassing U-Net by 2.21 and 1.53 percentage points, respectively.