Multimodal land use/cover classification using optical and synthetic aperture radar (SAR) images has attracted significant attention because the unique radiation and geometric characteristics of these images provide complementary information regarding land properties. However, the significant differences between these modalities create a large semantic gap, posing challenges for effective feature fusion in multimodal learning. Moreover, missing modalities often occur in practical applications due to weather constraints or sensor malfunctions, posing challenges to achieving high performance in cross-modal learning. In this study, we proposed a multimodal online knowledge distillation (MMOKD) framework, designed for land use/cover classification of optical and SAR images using either full or missing modalities. This framework trains one modality-fusion network alongside two modality-specific networks in an end-to-end manner, facilitating both multimodal and cross-modal learning. More specifically, we developed a multimodal feature fusion (MFF) module for integrating heterogeneous features and a single-modal feature generation (SFG) module for encapsulating cross-modal complementary information. In addition, we proposed the joint distillation with multitype fusion knowledge (JD-MFK) method, guiding the modality-specific student networks to comprehensively learn the modality-fusion teacher network. Notably, we adopted an online distillation strategy for real-time feedback and synchronous updates of both modality-fusion and modality-specific networks. Finally, we conducted extensive experiments on two multimodal land use/classification datasets with advanced multimodal fusion, cross-modal distillation, and specific baseline networks for comparison. The results demonstrate the effectiveness of the proposed MMODD, which not only outperforms the other networks in both full- and missing-modality scenarios but also significantly improves model training efficiency.

The accuracy of road damage detection models based on UAV remote sensing images is generally low, mainly due to the challenges posed by the complex background of road damage, diverse forms, and necessary computational requirements. To tackle the issue, this paper presents CSGEH-YOLO, a lightweight model tailored for UAV-based road damage detection in intricate environments. (1) The star operation from StarNet is integrated into the C2f backbone network, enhancing its capacity to capture intricate details in complex scenes. Moreover, the CAA attention mechanism is employed to strengthen the model’s global feature extraction abilities; (2) a cross-scale feature fusion strategy known as GFPN is developed to tackle the problem of diverse target scales in road damage detection; (3) to reduce computational resource consumption, a lightweight detection head called EP-Detect has been specifically designed to decrease the model’s computational complexity and the number of parameters; and (4) the model’s localization capability for road damage targets is enhanced by integrating an optimized regression loss function called WiseIoUv3. Experimental findings indicate that the CSGEH-YOLO algorithm surpasses the baseline YOLOv8s, achieving a 3.1% improvement in mAP. It also reduces model parameters by 4% and computational complexity to 78%. In contrast to alternative methods, the model proposed in this paper significantly reduces computational complexity while improving accuracy. It offers robust support for deploying UAV-based road damage detection models.

In order to improve the fidelity of SAR image simulation for ship target, we propose a novel ship target SAR dataset generation method based on ray tracing including multiple scattering. The method integrates the SAR spatial traversal mode and ray casting theory for discriminating the radar beam radiation areas of ship target and modeling electromagnetic (EM) wave transmissions under the geometric constraints of Range Doppler (RD) imaging geometry. Based on the simulation of EM waves transmission, the ray tracing algorithm and the physical illumination model are combined to simulate the multiple backscattering of EM waves, and the multiple backscattering amplitude and phase are recorded in real time. Then, the total backscattering field of the distributed ship target can be generated by the vector superposition of the multiple backscattering subfields of the EM waves transmitted towards the beam radiation area. Finally, the raw echo signal of the ship target is processed by RD imaging algorithm, then the corresponding SAR simulated image is obtained. We select some ship targets to simulate SAR images, and then compare the simulated results with the real SAR images and the simulated results of the mainstream commercial simulation software to verify the effectiveness of the proposed method.

In order to achieve an effective balance between SAR image simulation fidelity and efficiency, we proposed a ray-tracing-assisted SAR image simulation method under range doppler (RD) imaging geometry. This method utilizes the spatial traversal mode of RD imaging geometry to transmit discrete electromagnetic (EM) waves into the SAR radiation area and follows the Nyquist sampling law to set the density of transmitted EM waves to effectively identify the beam radiation area. The ray-tracing algorithm is used to obtain the backscatter amplitude and real-time slant range of the transmitted EM wave, which can effectively record the multiple backscattering among the components of the distributed target so that the backscattering subfields of each component can be correlated. According to the RD condition equation, the backscattering amplitude is assigned to the corresponding range gate, and the three-dimensional (3D) target is mapped into the two-dimensional (2D) SAR slant-range coordinate system, and the SAR target simulated image is directly obtained. Finally, the simulation images of the proposed method are compared qualitatively and quantitatively with those obtained by commercial simulation software, and the effectiveness of the proposed method is verified.

To achieve high-fidelity synthetic aperture radar (SAR) simulated images of typical target, we propose an echo-based SAR imaging simulation method assisted by ray tracing. The method incorporates multiple scattering mechanisms and establishes a backscattering model under the far-field plane wave assumption. A ray tracing algorithm is employed to capture the amplitude and phase of multiple backscattering components from numerous microfacets distributed across the target surface. The total backscattering field is then computed through vector superposition. Based on the principle of azimuthal synthetic aperture, we account for the real-time relative motion between the sensor and the target. A global phase correction is performed by leveraging the slant range history. This process transforms the local far-field scattering phase, which is referenced to the target modeling center, into a globally coherent scattering field. Then, the target’s global backscattering field is accurately reconstructed and serves as the raw echo signal. The range-Doppler (RD) imaging algorithm is subsequently applied to focus the simulated echo signal and generate the final SAR image. We select a typical aircraft carrier target for testing evaluation. The target is placed in the beam radiation scene and subjected to multi-angle rotation to produce corresponding multi-angle simulated SAR images. Qualitative and quantitative comparisons are conducted against both the RaySAR commercial software and real SAR image to verify the effectiveness of the proposed method.