Changtao He

Laser-induced damage of sol–gel SiO₂ antireflection coatings remains a key reliability issue in high-power laser systems because porous networks, residual hydroxyl groups, and defect-related absorption centers can trigger localized heating and stress concentration under nanosecond irradiation. In this work, continuous-wave CO₂ laser conditioning was used as a localized post-treatment method to regulate the microstructure of sol–gel SiO₂ coatings on fused silica substrates. The revised manuscript clarifies the processing window, scanning parameters, laser damage testing protocol, and the sample-specific nature of the reported LIDT values. Laser conditioning induces partial densification of the porous coating, dehydration of Si-OH groups, relaxation of the Si-O-Si network, and enhancement of mechanical properties. Under the optimized conditioning condition, the surface roughness decreases from 14.08 nm to 9.76 nm, and the LIDT at 1064 nm increases from 4.8 J/cm² to 7.0 J/cm². The LIDT values are discussed as a relative microstructure–property comparison for the present coating system rather than as the upper technological limit of sol–gel silica coatings. Combined FTIR analysis, thermal simulation, morphology observation, and damage probability analysis indicate that the improvement originates from the combined effects of reduced defect absorption, moderated porosity, improved heat dissipation, and enhanced resistance to thermally induced cracking. The results provide a mechanism-guided strategy for using CO₂ laser conditioning to tune sol–gel silica coatings while also identifying the need for further validation on higher-LIDT coatings and at application-relevant wavelengths.

Currently, an increasing number of convolutional neural networks (CNNs) focus specifically on capturing contextual features (con. feat) to improve performance in semantic segmentation tasks. However, high-level con. feat are biased towards encoding features of large objects, disregard spatial details, and have a limited capacity to discriminate between easily confused classes (e.g., trees and grasses). As a result, we incorporate low-level features (low. feat) and class-specific discriminative features (dis. feat) to boost model performance further, with low. feat helping the model in recovering spatial information and dis. feat effectively reducing class confusion during segmentation. To this end, we propose a novel deep multi-feature learning framework for the semantic segmentation of VHR RSIs, dubbed MFNet. The proposed MFNet adopts a multi-feature learning mechanism to learn more complete features, including con. feat, low. feat, and dis. feat. More specifically, aside from a widely used context aggregation module for capturing con. feat, we additionally append two branches for learning low. feat and dis. feat. One focuses on learning low. feat at a shallow layer in the backbone network through local contrast processing, while the other groups con. feat and then optimizes each class individually to generate dis. feat with better inter-class discriminative capability. Extensive quantitative and qualitative evaluations demonstrate that the proposed MFNet outperforms most state-of-the-art models on the ISPRS Vaihingen and Potsdam datasets. In particular, thanks to the mechanism of multi-feature learning, our model achieves an overall accuracy score of 91.91% on the Potsdam test set with VGG16 as a backbone, performing favorably against advanced models with ResNet101.