Optical and Synthetic Aperture Radar (SAR) imagery offers a wealth of complementary information on a given target, attributable to the distinct imaging modalities of each component image type. Thus, multimodal remote sensing data have been widely used to improve land cover classification. However, fully integrating optical and SAR image data is not straightforward due to the distinct distributions of their features. To this end, we propose a land cover classification network based on multimodal feature fusion, i.e., MFFnet. We adopt a dual-stream network to extract features from SAR and optical images, where a ResNet network is utilized to extract deep features from optical images and PidiNet is employed to extract edge features from SAR. Simultaneously, the iAFF feature fusion module is used to facilitate data interactions between multimodal data for both low- and high-level features. Additionally, to enhance global feature dependency, the ASPP module is employed to handle the interactions between high-level features. The processed high-level features extracted from the dual-stream encoder are fused with low-level features and inputted into the decoder to restore the dimensional feature maps, generating predicted images. Comprehensive evaluations demonstrate that MFFnet achieves excellent performance in both qualitative and quantitative assessments on the WHU-OPT-SAR dataset. Compared to the suboptimal results, our method improves the OA and Kappa metrics by 7.7% and 11.26% on the WHU-OPT-SAR dataset, respectively.

Metalens can achieve arbitrary light modulation by controlling the amplitude, phase, and polarization of the incident waves and have been applied across various fields. This paper presents a color router designed based on metalens, capable of effectively separating spectra from visible light to near-infrared light. Traditional design methods for meta-lenses require extensive simulations, making them time-consuming. In this study, we propose a deep learning network capable of forward prediction across a broad wavelength range, combined with a particle swarm optimization algorithm to design metalens efficiently. The simulation results align closely with theoretical predictions. The designed color router can simultaneously meet the theoretical transmission phase of the target spectra, specifically for red, green, blue, and near-infrared light, and focus them into designated areas. Notably, the optical efficiency of this design reaches 40%, significantly surpassing the efficiency of traditional color filters.

The threat posed by adversarial examples (AEs) to deep learning applications has garnered significant attention from the academic community. In response, various defense strategies have been proposed, including adversarial example detection. A range of detection algorithms has been developed to differentiate between benign samples and adversarial examples. However, the detection accuracy of these algorithms is significantly influenced by the characteristics of the adversarial attacks, such as attack type and intensity. Furthermore, the impact of image preprocessing on detection robustness—a common step before adversarial example generation—has been largely overlooked in prior research. To address these challenges, this paper introduces a novel adversarial example detection algorithm based on high-level feature differences (HFDs), which is specifically designed to improve robustness against both attacks and preprocessing operations. For each test image, a counterpart image with the same predicted label is randomly selected from the training dataset. The high-level features of both images are extracted using an encoder and compared through a similarity measurement model. If the feature similarity is low, the test image is classified as an adversarial example. The proposed method was evaluated for detection accuracy against four comparison methods, showing significant improvements over FS, DF, and MD, with a performance comparable to ESRM. Therefore, the subsequent robustness experiments focused exclusively on ESRM. Our results demonstrate that the proposed method exhibits superior robustness against preprocessing operations, such as downsampling and common corruptions, applied by attackers before generating adversarial examples. It is also applicable to various target models. By exploiting semantic conflicts in high-level features between clean and adversarial examples with the same predicted label, the method achieves high detection accuracy across diverse attack types while maintaining resilience to preprocessing, providing a valuable new perspective in the design of adversarial example detection algorithms.

At detection ranges on the order of kilometers, cat-eye echoes degrade into pixel-level spots and exhibit responses similar to those of compact, high-echo false targets, making it difficult for single-band systems to balance weak-target detection rates with false alarm control. To address this, this paper pairs spatially registered and radiometrically calibrated 808 nm and 905 nm images to form a dual-band detection pair, utilizing the cross-band response asymmetry caused by chromatic defocus for discrimination. Experiments show that cat-eye targets exhibit coupled asymmetry in intensity, scale, and energy distribution, whereas the false targets under test generally maintain approximate symmetry; furthermore, the dominant discriminative information shifts from scale differences to intensity differences as distance increases. Based on this pattern, this paper proposes a range-conditioned “detection–reclassification” framework that dynamically matches appropriate models based on target distance. In this framework, a structurally symmetric, two-branch, multi-scale attention-fused YOLO network is used for candidate target detection, while an explicit asymmetric feature classifier further processes challenging samples to retain weak targets and suppress false alarms. Field experiments conducted at distances ranging from 100 to 2100 m achieved an overall mAP50 of 0.907 and an overall mAP50–95 of 0.475, with the mAP50 improving by 17.7 and 10.8 percentage points, respectively, compared to the 808 nm and 905 nm single-band baselines. The results demonstrate that this framework can effectively balance the detection of weak targets at long ranges with the suppression of false targets under the evaluated conditions.