Yuanyi Fan

Light Detection and Ranging (LiDAR) has become a pivotal technology for non-cooperative target detection and characterization in space situational awareness, on-orbit servicing, and deep-space exploration, owing to its high spatial resolution and precise ranging capability. However, in complex dynamic environments, non-cooperative target detection is affected by multi-source interferences including temperature and vibration, as well as time-varying parameters such as distance and incidence angle. These factors jointly cause nonlinear fluctuations in laser echo intensity. Traditional geometric reconstruction methods rely on the ideal Lambertian surface assumption and cannot decouple these nonlinear modulation effects, leading to errors in material property inversion. For non-Lambertian surface laser echo signals, this paper proposes an instantaneous state sampling and parametric decoupling method. A Wendland C$^2$-Gaussian hybrid interpolation model is adopted to parametrically characterize the temperature-vibration coupling effect, and a momentum-enhanced stochastic Bregman proximal gradient (MSBPG) method is used for global optimization. The Cook-Torrance bidirectional reflectance distribution function is introduced to model the target surface, isolating the temperature-vibration coupling effect from time-varying spatial modulation. Experimental validation is conducted using a semi-physical simulation ground system with controllable temperature field and multi-axis vibration excitation platform. Results show that under compound interference of 353K thermal radiation and 25~Hz mechanical vibration, the MSBPG algorithm reduces the decoupling root mean square error (RMSE) by 45.2% compared with the Euclidean proximal method. Compared with the traditional Lambertian-based distance-angle intensity correction method, the proposed method improves target surface emissivity estimation accuracy by 32.5%. This method provides a high-precision and physically interpretable solution for non-cooperative target recognition and dynamic material characterization in harsh environments.

Recent advances in transformer architectures have demonstrated strong performance across diverse computer vision tasks, including multiview 3D reconstruction. However, when handling a large number of input views, conventional transformers face challenges of efficiency and representational complexity due to the abundance of image tokens and the richness of visual content. This results in high computational overhead and difficulty in modeling meaningful inter-view relationships. Existing approaches often mitigate these issues by reducing token counts or discarding cross-view attention, which typically degrades performance. To address these limitations, we propose CCTNet, a transformer architecture based on a divide-and-conquer strategy. At its core, we introduce the context coupling token (CCT) mechanism, which aggregates tokens from all views and partitions them into multiple groups. Each group integrates tokens sampled across all perspectives, yielding a holistic representation that preserves inter-view diversity. This design enables our encoder to capture inter-view dependencies through grouped attention, while standard self-attention layers retain intra-view feature learning. Furthermore, a proactive upsampling decoder is employed to efficiently generate accurate voxel outputs. Extensive experiments on the ShapeNet dataset demonstrate that CCTNet achieves an IoU of 0.778 with 20 views, matching or surpassing prior methods such as UMIFormer (0.778) and GARNet+ (0.742), while requiring significantly fewer computations (4.5 GFLOPs versus 11.7 GFLOPs for K-means clustering). These results highlight that CCTNet delivers both superior accuracy and efficiency for multiview 3D reconstruction.

Surface normals provide critical geometric information about object surfaces and support various computer vision tasks. However, estimating high-quality scene-level normals using single-polarization images remains challenging. Existing SfP methods are predominantly data-driven, directly learning implicit mappings from polarization images to normals but neglecting the diverse prior information embedded in the polarization imaging process, which makes it difficult to effectively address challenges such as azimuth ambiguity, observation noise, and non-orthographic projection. To address this, this paper proposes a physically enhanced Shape from Polarization (SfP) framework built upon a more comprehensive imaging model, which integrates multi-head self-attention and multiple physical priors to achieve high-quality scene-level normal estimation. We introduce three key priors into our network: a reflection type prior based on the imaging model, an angle of polarization (AoP) confidence metric representing the reliability of polarization information, and a degree of polarization (DoP) soft boundary capturing the geometric features of different objects. Furthermore, to handle polarization imaging under non-orthogonal projections, we incorporate encoded light directions into our model in order to represent the geometric relationship connecting the real world to the image. The experimental results demonstrate that our method outperforms existing scene-level SfP approaches. Benefiting from our reversible polarization physical model, it generalizes well to challenging outdoor large-scale scenes.

Images of underwater environments suffer from contrast degradation, reduced clarity, and information attenuation. The traditional method is the global estimate of polarization. However, targets in water often have complex polarization properties. For low polarization regions, since the polarization is similar to the polarization of background, it is difficult to distinguish between target and non-targeted regions when using traditional methods. Therefore, this paper proposes a joint evaluation and partition fusion method. First, we use histogram stretching methods for preprocessing two polarized orthogonal images, which increases the image contrast and enhances the image detail information. Then, the target is partitioned according to the values of each pixel point of the polarization image, and the low and high polarization target regions are extracted based on polarization values. To address the practical problem, the low polarization region is recovered using the polarization difference method, and the high polarization region is recovered using the joint estimation of multiple optimization metrics. Finally, the low polarization and the high polarization regions are fused. Subjectively, the experimental results as a whole have been fully restored, and the information has been retained completely. Our method can fully recover the low polarization region, effectively remove the scattering effect and increase an image’s contrast. Objectively, the results of the experimental evaluation indexes, EME, Entropy, and Contrast, show that our method performs significantly better than the other methods, which confirms the feasibility of this paper’s algorithm for application in specific underwater scenarios.

The study of pixelated metamaterials that integrate both the functions of linear and circular polarization filters is rapidly growing due to the need for full-Stokes polarization imaging. However, there is a lack of large-area, ultracompact pixelated full-Stokes metamaterials with excellent performance, especially circular polarization filters with a high extinction ratio, a broad operating bandwidth, and a low-cost, high-quality, efficient manufacturing process, which limits the practical applications of pixelated full-Stokes metamaterials. In this study, we propose a universal design and fabrication scheme for large-area, ultracompact pixelated aluminum wire-grid-based metamaterials used in Vis-NIR full-Stokes polarization imaging. The aluminum wire-grid was designed as a linear polarization filter with an average linear polarization extinction ratio of 36,000 and a circular polarization filter with an average circular polarization extinction ratio of 110 in Vis-NIR. A large-area, ultracompact 320×320 pixelated aluminum wire-grid-based full-Stokes metamaterial was fabricated using nanoimprint lithography and nano transfer printing with the advantages of low cost and high efficiency. This metamaterial was used to achieve full-Stokes polarization imaging with errors within 8.77%, 12.58%, 14.04%, and 25.96% for Stokes parameters S₀, S₁, S₂, and S₃, respectively. The inversion errors of the compensated Stokes parameters can be reduced to 0.21%, 0.21%, 0.42%, and 1.96%, respectively.

Chiral metamaterials are extensively applied in the fields of photoelectric detection, biomedical diagnostics and micro-nano polarization imaging. Currently, single-layer chiral metamaterials are unfortunately limited by several issues, such as a weaker circular polarization extinction ratio and circular polarization transmittance difference. To tackle these issues, a single-layer transmissive chiral plasma metasurface (SCPMs) suitable for visible wavelength is proposed in this paper. Its basic unit is composed of double orthogonal rectangular slots and a spatial π/4 inclined arrangement of the rectangular slot to constitute a chiral structure. Each rectangular slot structure has characteristics that enable the SCPMs to easily achieve a high circular polarization extinction ratio and strong circular polarization transmittance difference. Both the circular polarization extinction ratio and circular polarization transmittance difference of the SCPMs reach over 1000 and 0.28 at a wavelength of 532 nm, respectively. In addition, the SCPMs is fabricated via the thermally evaporated deposition technique and focused ion beam system. This compact structure coupled with a simple process and excellent properties enhances its applicability for the control and detection of polarization, especially during integration with linear polarizers, to achieve the fabrication of a division-of-focal-plane full-Stokes polarimeter.

A universal design and fabrication scheme is proposed for a cascaded aluminum wire‐grid‐based vis–NIR circular polarizer. The aluminum wire‐grid is designed as a broadband circular‐linear polarization converter and a broadband linear polarizer, which is further cascaded to form a circular polarizer with strong circular polarization transmittance difference (CPTD) and circular polarization extinction ratio (CPER) in vis–NIR. By increasing the distance between the circular‐linear polarization converter and the linear polarizer, the redshift of the CPTD and CPER peaks is realized and used to design a novel circular polarizer to work better in the operating wavelength. Anisotropic thin‐film interference is revealed to be responsible for the redshift phenomenon. Furthermore, the circular polarizer is fabricated by nano imprint lithography and nano transfer printing with the merits of relatively low cost, and high production efficiency. With such flexibility in the design and fabrication, cascaded aluminum wire‐grid may serve as a tunable compact circular polarizer, and it is also easy to integrate with the wire‐grid‐based linear polarizer to realize the fabrication of the division‐of‐focal‐plane full Stokes polarimeter.