Abstract: Targeting ash deposits on the platen superheater of a 660 MW boiler firing Xinjiang high-alkali coal, a high-temperature visual imaging system was established to acquire real-time ash deposition images. Ash deposition images were preprocessed through grayscale conversion, grayscale linearization, and Gaussian filtering, and quantitatively evaluated via peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM). An enhanced image dataset was constructed through image augmentation techniques. Leveraging the correlation between deposit thickness and thermal resistance, a convolutional neural network (CNN)-based monitoring model was developed for ash deposit identification. Results show that the CNN method achieves effective real-time monitoring of ash deposit status on operating platen superheaters, with accuracy, recall, and F_score reaching 98.8%, 98.2%, and 98.8% respectively.