In computer vision and computer graphics, 3D reconstruction aims to generate high-fidelity digital models of real objects and scenes from 2D images, videos, or other sensor data, and has long attracted attention from both academia and industry. In recent years, deep learning has continuously improved reconstruction accuracy and efficiency, enabling applications such as relighting, digital twins, and virtual reality. With the advent of Neural Radiance Fields (NeRF), neural-field-based reconstruction has become increasingly mainstream; inverse rendering has been incorporated to enhance appearance disentanglement, forming a more complete reconstruction framework. Motivated by a rapidly growing yet fragmented literature, this survey focused on static objects and scenes and organized neural-field-based inverse-rendering 3D reconstruction into two components: neural-field surface reconstruction, and neural-field material and illumination estimation. For each component, core research questions were distilled and representative methods were summarized following a “research problem-technical solution” perspective. Their key principles and derivations were analyzed to inform future algorithm design. In addition, commonly used datasets and evaluation metrics for both directions were compiled, and their practical details and applicability were discussed. Finally, open challenges and promising future directions were outlined.

Control variate techniques constitute a class of classical and effective variance reduction methods and have demonstrated significant advantages in a wide range of Monte Carlo estimation problems in computer graphics rendering in recent years. To address the pervasive issues of high variance and slow convergence in rendering computations, this survey systematically reviewed the theoretical foundations and recent advances of control variate techniques in the rendering domain. By introducing auxiliary functions highly correlated with the target integrand and whose expectations are analytically tractable or accurately estimable, control variates transformed the original estimators into unbiased forms with substantially reduced variance. Focusing on key computational components such as path tracing, many light sampling, free-path sampling, transmittance estimation, and image reconstruction, this survey categorized and summarized representative control variate-based methods across surface rendering, volume rendering, re-rendering, inverse rendering, and gradient-domain rendering. Furthermore, the characteristics of different approaches were analyzed in terms of control variate representations, construction strategies, applicable scenarios, and quality improvement performance, and major challenges and future research directions were discussed.

The recognition of Tibetan characters in complex backgrounds is one of the important research topics in the field of Tibetan OCR (Optical character Recognition). However, traditional deep learning models often have excessive parameter counts and are difficult to deploy, affecting the efficiency of Tibetan character recognition. To address the above problems, a Tibetan character recognition model for complex backgrounds based on the lightweight MobileNetV3-Global was proposed. The global attention mechanism (Global Context Block, GCB) was introduced into the last layer of the baseline model MobileNetV3. While maintaining the computational complexity of the network model, the global attention mechanism was used to perform global modeling on the line-text images of Tibetan characters in complex backgrounds, thereby facilitating feature extraction. Experimental results demonstrated that the improved lightweight MobileNetV3-Global model achieved high accuracy for Tibetan character recognition in complex backgrounds. In this study, tests were conducted on two types of datasets, namely natural scenes and streaming media. The accuracy rates of the model on these datasets reached 97.40% and 96.15%, respectively, and the model size was only 6.05 MB, indicating high practicality and deployment potential.

Traditional 3D Gaussian Splatting (3DGS) relies on time-consuming and fragile Structure-from-Motion (SfM) preprocessing, leading to significant application bottlenecks. Although recent approaches leverage pretrained Multi-View Stereo (MVS) models to bypass SfM, they suffer from high computational cost and memory overhead, rendering them unsuitable for dense-view or large-scale reconstruction. To address these issues, an efficient 3DGS framework based on VGGT preprocessing and adaptive voxelization was proposed, which reduced dependence on input views, camera poses, and training resources, enabling end-to-end scene reconstruction within minutes. Firstly, VGGT required only input images and was used to infer camera poses and initial dense scene point clouds within seconds. Then, a depth-refined point-cloud reconstruction module was designed to enhance geometric completeness, boundary sharpness, and fine-grained realism of the initial point cloud. Then, an adaptive voxelization strategy guided by multi-dimensional image saliency was introduced to prune redundant Gaussian primitives during training, significantly reducing memory usage. Finally, confidence-aware depth regularization was combined with multi-view geometric consistency constraint to compensate for rendering quality degradation after voxelization, achieving an optimal balance between compression efficiency and visual fidelity. Experimental results demonstrate that, compared with the current advanced methods, the proposed method achieves consistent improvements in PSNR, SSIM, and LPIPS on the TNT (sparse and dense views) and Mip-NeRF360 (sparse views) datasets. Meanwhile, the number of Gaussian primitives is reduced by approximately 25% and 67% in sparse and dense scenes, respectively, enabling fast and high-quality scene reconstruction.

In game development, Procedural Content Generation (PCG) effectively reduces development costs and enhances the diversity of level design. Recent advances in integrating PCG with machine learning have demonstrated significant progress in game level generation. However, existing methods still exhibit limitations in generation controllability and adaptability to complex design intentions. To address these limitations, the LlaMario model was proposed to convert Mario game levels into character matrices for training with Large Language Models (LLMs), and then transformed the generated symbolic sequences into playable Mario levels. Specifically, we adopted the Llama 3.1-8B-instruct model combined with the efficient language model to fine-tune the Unsloth framework, empowering the model with enhanced level comprehension and generation capabilities. An instruction set of 40 000 entries derived from the Gemini model’s comprehension of level data was constructed and employed together with the data-efficient Alpaca instruction-tuning strategy for dataset construction; the resulting LlaMario trained on this dataset generated highly playable Mario levels from natural language descriptions. Experimental results validated that the proposed model achieved exceptional performance in level logical coherence, playability, complexity, and generation quality, successfully producing user-intended Mario game levels. Furthermore, the proposed framework demonstrated generality and was extended to other tile-based level generation tasks.

With the rapid development of artificial intelligence, deepfake technology, such as generative adversarial networks and diffusion models, has advanced to generate highly realistic images, posing substantial challenges to traditional image forgery detection methods. Existing methods often treat forgery detection and forgery localization as separate tasks, which results in insufficient feature information sharing and weak synergy between the two tasks, therefore limiting the detection accuracy and localization precision. To address this issue, a novel localization-branch- assisted image forgery detection framework was proposed, and multi-task collaborative optimization was used to enhance detection performance. The proposed model included two key innovative modules. First, a Dual-Domain Feature Enhancement Module was designed to combine color distribution anomalies with high-frequency distortion features, so that forgery traces could be jointly captured and discriminative capability enhanced, addressing the limited adaptability of a single feature in complex forgery scenarios. Second, a Median-Enhanced Interaction Module was introduced to enhance interaction between the detection and localization branches through multi-scale feature fusion and a combined channel-spatial attention mechanism, effectively mitigating the task-isolation problem. An end-to-end joint training framework with a multi-task loss function was adopted to further reinforce the synergy between detection and localization tasks. Extensive experiments were conducted on datasets such as Columbia, CASIA, and NIST16, and the results showed that the proposed method outperformed existing comparison approaches in both detection and localization tasks, demonstrating the effectiveness and generalization capability in complex forgery scenarios.

To address insufficient spatiotemporal feature coupling and the inadequate utilization of multimodal information in dance motion recognition, a multi-modal fusion Beat STMAN (Beat-guided Spatio-Temporal Multimodal AMAA-Net Network) network recognition method was proposed to improve the accuracy of dance action recognition. Firstly, this method is based on the skeleton based ST-GCN (Spatial Temporal Graph Convolutional Network) model to construct a spatiotemporal convolutional skeleton network. To cope with the unfavorable factors of continuous and varied dance movements and partially obscured movements, a dynamic adjacency matrix was integrated with a multi-head spatial attention mechanism to automatically capture global human-body topology parameters. Secondly, an audio stream information feature extraction alignment fusion method was proposed to obtain beat timestamp pulse sequences, and a Transformer multi-head attention mechanism was used to design a cross-modal fusion module AMAA-Net, which achieved multimodal feature fusion through a resistance game mechanism, effectively alleviating insufficient model feature fusion. Finally, the Beat-STMAN was evaluated on a publicly available dance dataset to verify its effectveness. Experimental results showed that for the Thomas spin movement, the recognition rate of the proposed model achieved 14.7% higher than that of the ST-GCN model, demonstration significantly improved robustness in occluded scenarios. Furthermore, ablation experiments verified that the integration of the dynamic adjacency matrix, multi-head attention mechanism, and cross-modal attention mechanism could effectively fuse audio-action correlation features, with a cross-modal contribution rate reaching 5.3%. This effectively improved the model’s Top-1 accuracy, thereby enhancing the model’s prediction precision and providing a multimodal technical implementation path for dance-teaching evaluation and immersive interaction.

To address the problems of low small-target segmentation accuracy, insufficient cross-modal feature extraction and unsatisfactory fusion effect in multimodal remote sensing image semantic segmentation models, a dual-branch multimodal network Segment Anything Model Low-Frequency Fusion (SAMLoF), based on adapter optimization and low-frequency information extraction and fusion was proposed. By efficiently fine-tuning the parameters of the Segment Anything Model image encoder through the proposed joint adapter structure, the model could learn specific knowledge in the field of remote sensing while maintaining its general capabilities. The specific low-frequency information input modules for different modal images were designed to provide more large-scale environmental structures and contextual information for the model through the Fast Fourier Transform. An efficient feature fusion module based on attention was proposed to promote the full fusion of relevant information in different modal features. SAMLoF was evaluated and compared with 15 methods on the Vaihingen and Potsdam datasets. Both mF1 and mIoU metrics of SAMLoF reached the optimal values. The results showed that SAMLoF could effectively extract and fuse different modal features in remote sensing images, especially generating accurate and smooth boundary contours for complex small target objects.

Monocular depth estimation has attracted sustained attention due to its broad application prospects in autonomous driving, 3D reconstruction, and related fields. However, existing methods still exhibit deficiencies in multi-scale feature fusion and depth modeling, which makes it difficult to simultaneously capture fine-grained local geometric details and maintain global structural consistency, and have limited adaptability to depth scale distributions across different scenes. To address these issues, a monocular depth estimation framework integrating hierarchical dual-stream attention and adaptive depth discretization was developed. In the encoding stage, a Swin Transformer was employed to construct pyramid multi-scale feature representations, enhancing the joint modeling of local and global information. In the decoding stage, a hierarchical dual-stream attention fusion network was designed to model local detail perception and global contextual semantics in parallel during progressive reconstruction, in which adaptive feature fusion was achieved through dynamic weight modulation and cross-attention mechanisms. Meanwhile, a depth recovery module was introduced to formulate depth estimation as a joint classification-regression task. By predicting a discrete depth distribution and adaptively learning the bin centers, continuous depth values were generated via probability-weighted aggregation, which ensured the continuity of depth prediction while effectively maintaining the consistency of depth relationships within the scene. Experimental results demonstrated that the proposed method achieved state-of-the-art performance on the KITTI dataset, with an AbsRel of 0.048, a SqRel of 0.147, a log10 error of 0.020, and a δ? accuracy of 0.980, and exhibited favorable generalization capability on the NYU Depth V2 dataset, validating its effectiveness and robustness under complex scenes and multi-scale depth distributions.

3D Gaussian Splatting (3DGS) technology has demonstrated exceptional performance in the field of novel view synthesis. However, when processing complex scenes with abundant details, its conventional densification strategy is highly susceptible to the over-reconstruction phenomenon. This leads to a small number of large-volume Gaussian primitives fitting extensive areas, resulting in noticeable blurriness in rendered images and a severe loss of high-frequency texture details. To overcome this issue, a novel Frequency Intensity Gaussian Splatting (FIGS) algorithm was proposed. First, starting from local image frequency-domain characteristics, a “Frequency Intensity” metric was defined to precisely quantify the richness of high-frequency details within local spatial regions, establishing corresponding frequency intensity maps. Second, a frequency-intensity regularization mechanism was designed. By incorporating the discrepancy between the frequency-intensity maps of the rendered images and the ground truth into the loss function, this mechanism drove the algorithm to adaptively allocate more Gaussian primitives to over-reconstructed regions densely populated with high-frequency details. Finally, a confidence-based adaptive Gaussian evolution strategy was introduced. Leveraging multi-view consistency priors to construct a geometric confidence factor, this strategy achieved robust Gaussian evolution through a joint mechanism of weighted densification and active pruning, thereby dynamically optimizing Gaussian allocation across the entire scene. Extensive qualitative and quantitative experiments conducted on multiple real-world benchmark datasets, including Mip-NeRF360 and Tanks & Temples, demonstrated that the proposed algorithm achieved highly competitive results in evaluation metrics such as Structural Similarity (SSIM) and Learned Perceptual Image Patch Similarity (LPIPS), effectively reducing the proportion of over-reconstructed regions. The Frequency Intensity Gaussian Splatting algorithm could effectively alleviate blurriness during the reconstruction process and scientifically orchestrate the spatial distribution of Gaussian primitives. It exhibited outstanding performance in detail fidelity and overall visual quality for novel view synthesis, providing a robust new solution for high-precision 3D reconstruction of complex real-world scenes.

To address insufficient texture detail restoration and inadequate inter-frame consistency modeling in existing video dehazing methods, a progressive spatiotemporal detail-enhanced video dehazing algorithm (Progressive Spatiotemporal Detail Enhancement Network，PSTD-net) was proposed, comprising two modules: a spatial detail enhancement and a cross-frame temporal modeling. First, to alleviate insufficient high-frequency texture modeling in existing methods, a progressive detail-enhancement encoder was proposed to effectively extract and restore texture details damaged by haze. Second, to improve the temporal consistency between video frames, a cross-frame detai- enhancement module was designed to model inter-frame dependencies through a spatiotemporal attention mechanism, enhancing image details while suppressing artifact generation. Experiments conducted on multiple synthetic and real hazy video datasets showed that PSTD-net improved both dehazing effect and visual consistency, providing a new solution for video dehazing tasks.

To address automatic generation of structured hexahedral meshes, an automatic hexahedral mesh generation algorithm based on model-domain segmentation was proposed. This method constructed an attribute-adjacency graph from geometric attribute of the model to extract domain features and generate partitioning surfaces, thereby decomposing a model domain into multiple simpler subdomains. The subdomains were then automatically classified and matched based on their features, and predefined subdomain mesh templates were applied to generate subdomain meshes. Finally, the subdomain meshes were merged to obtain an all-hexahedral mesh for the entire model. By combining feature-driven domain segmentation with template-based mesh generation, the proposed method reduced manual intervention while ensuring both topological consistency and geometric feature preservation, thereby improving the stability and repeatability of mesh generation. In the template-generation stage, feature-preserving parameterized mapping was combined with topology-constrained sweeping to generate the hexahedral meshes, forming an automated pipeline from domain segmentation to template generation and mesh merging. The method was applied to meshing typical mechanical models, providing a reference for digital simulation and analysis.

Current Immersive Physics Experiments (IPE) generally lack haptic feedback, leading to insufficient interaction realism and negatively affecting students’ experimental operation experience and learning outcomes. To address this issue, a semi-physical interaction technology integrating active and passive haptic feedback was proposed. Drawing from the concept of semi-physical simulation, the proposed technology constructed an experimental system with synchronized visual-haptic feedback by integrating 3D-printed physical entities, multi-type sensors (tension, pressure, temperature, etc.), and actuators. Specifically, a semi-physical interaction device was designed for the coupled spring-mass oscillator experiment integrating a puller, button, and knob; a device for the three gas laws experiment based on a piston and temperature control module; and a device for the buoyancy experiment composed of digital servos and a jet device. A controlled experiment involving 32 university students was conducted, where the experimental group used semi-physical interaction and the control group used gesture interaction. Evaluation was performed using the NASA-TLX (NASA-Task Load Index) subjective scale combined with 64-channel EEG signals. Subjective results indicated that the post-test total cognitive load of the experimental group was 55.27, which was significantly lower than that of the control group (60.13, p=0.041), with marked improvements in physical demand and frustration levels. Neurophysiological results demonstrated that the average power spectral densities of θ, α, and β bands in the frontal (Fz) and occipital (Oz) lobes of the experimental group showed a downward trend and were significantly lower than in the control group, indicating reduced pressure on neural resource allocation. The study confirmed that compared with traditional gesture interaction, semi-physical interaction effectively reduced students’ cognitive load in IPE, optimized the efficiency of cognitive resource allocation, and facilitated easier mastery of relevant physics knowledge.

To address the subjectivity and poor quantifiability of post-stroke upper-limb functional assessment, a digital-twin-based evaluation method targeted at the Fugl-Meyer Assessment of the Upper Extremity (FMA-UE) was proposed and validated in a controlled study. PN3 sensors were used to capture pose data and a Dynamic Fusion Rating Algorithm (DFRA) was constructed to achieve item-level alignment with FMA-UE, outputting individual scores and identifying deficits. In a controlled comparative study with 30 participants, the system improved efficiency by approximately 40% relative to manual assessment. System scores were statistically equivalent to therapist scores within a ±3-point equivalence margin under the Two One-Sided tests (TOST) procedure, with overall errors of MAE=1.97 points and RMSE=2.14 points. Deficit identification achieved Top-1 agreement of 88.0% with therapists across the five FMA-UE domains. The proposed DFRA-plus-digital-twin approach enabled item-level automated scoring and deficit visualization while preserving clinical interpretability, markedly improving efficiency over traditional assessment and demonstrating practical deployability in engineering settings.

To address the prevalent noise interference and missing-marker problem during the collection and transmission of Motion Capture (MoCap) data, a recovery model combining the Zero-norm sparsity and Temporal Differential Low-rank minimization (ZTDL) was proposed. Firstly, a temporal difference low-rank regularization term was introduced to capture the global low-rank property and temporal smoothness of MoCap data. Besides, the l₀ norm and the Frobenius norm were employed to characterize sparse missing noise and additive Gaussian noise. Secondly, the non-convex recovery model was transformed into the optimization problem involving a binary mask matrix by exploiting the properties of the l₀ norm. This model enabled the simultaneous estimation of missing regions and the restoration of MoCap data. The optimization problem was efficiently solved using the Alternating Direction Method of Multipliers (ADMM) framework sand the (inverse) Discrete Cosine Transform (DCT/IDCT). The algorithm was rigorously proven to converge to a local minimum in a coordinate-wise manner in theory. Finally, extensive comparative experiments were conducted on the benchmark CMU and HDM05 datasets. The ZTDL algorithm was evaluated against a range of classical methods such as TSMC, TRNN, IRNN-Lp, and TSPN, as well as deep learning approaches. The restoration results, including recovery error and visual effect, demonstrated the significant superiority of ZTDL in both missing-region estimation and corrupted-data restoration.

To address the limitations of current kernels in solid representation and the efficiency bottlenecks in complex solid modeling, a manifold/non-manifold unified solid representation method based on the manifold assembly concept was proposed. By interpreting non-manifold solids as a union of finite manifold subsets, this approach overcome the limitation of traditional approaches, which relied on non-manifold entities or non-manifold constraints to represent non-manifold solids. The construction logic of the solid representation data structure was simplified without compromising the range of solid representations. This laid the foundation for improving the efficiency and versatility of manifold/non-manifold modeling algorithms. Furthermore, a bidirectional and flexible topological linking mechanism, along with a hybrid topological-link maintenance strategy, was proposed. While ensuring topological consistency, the proposed method improved the topological retrieval efficiency of complex models (containing topological entities at the 10⁴ level) by nearly 90% compared with OCC, approaching the performance of leading international commercial kernels such as CGM. This provided technical support for further enhancing system modeling efficiency.

Sustainable development serves as a core approach to addressing global environmental and energy challenges, while the construction industry has gradually implemented green construction practices, but the inherent connection between construction environment characteristics and human subjective perception has not yet been systematically explored. This gap results in a lack of targeted theoretical support for the optimization of green construction in terms of safety, aesthetics, and efficiency. To clarify the laws governing human perception of construction environments and thereby guide the optimization of green construction practices, this research collected preference data from participants regarding randomly matched combinations of two typical construction scenario images through an online crowdsourcing platform. The Microsoft TrueSkill system was employed to quantitatively rank the perceptual preference results. For the key visual elements in the construction scenarios, the Region of Interest (ROI)-based analysis method was used for feature extraction and interpretation. Meanwhile, several Convolutional Neural Networks (CNN) were selected and trained to enable the automated prediction of perceptual quality. The results indicate that there is a significant statistical correlation between the subjective perception of construction sites in the perceptual dimensions mentioned above and specific visual features: the stacking neatness of on-site building materials, ground cleanliness and equipment usage affect the perception about safety, aesthetics and efficiency, respectively. Due to the fact that the first two types of visual features are both external manifestations of the construction site orderiness, the perception of safety and aesthetics exhibits a strong positive correlation, while no correlation was observed between efficiency perception and the other two types of perception. An innovative construction environment evaluation framework was proposed, integrating crowdsourcing technology with deep learning. The consistency of visual perception of construction scenarios across different groups was clarified, and the specific visual standards required for civilized construction sites were clarified, providing a benchmark foundation for the establishment of an automated evaluation system for the perceptual quality of construction environments.

In high-precision manufacturing domains such as aerospace, the assembly accuracy of complex products fundamentally determines their core performance and operational reliability. Traditional assemblyerror modeling methods exhibit significant limitations in capturing the nonlinear behaviors associated with “geometry-mechanics” coupling—particularly in accurately describing contact deformation and dynamic coupling effects, which are critical for high-precision applications. To address these challenges, an innovative Jacobian-Torsor Embedded Neural Network (JENN) approach was proposed. Physical constraints derived from Jacobian-Torsor theory were embedded into the neural network architecture, establishing a joint optimization mechanism that simultaneously minimized the “Jacobian-Torsor loss function” and the network’s prediction error, enabling precise modeling of complex nonlinear mechanisms such as contact deformation and dynamic coupling. The proposed method leveraged the multilayer nonlinear mapping capability of neural networks to capture latent physical correlations, while using kinematic constraints from Jacobian-Torsor theory to regularize network training. As a result, the model’s generalization ability and physical interpretability were significantly enhanced, particularly in small-sample scenarios. Experimental validation on a dual-axis turntable used azimuth and elevation angles as inputs to predict spatial pointing errors, demonstrating substantial improvements: compared with conventional networks without Jacobian embedding, prediction accuracy improved by 46.51%, 37.44%, and 50.86% in the x, y, and z directions, respectively. Five-fold cross-validation further showed that the proposed method consistently outperformed the conventional approach across four key metrics (MSE, RMSE, MAE, and R2), yielding lower prediction errors as well as markedly enhanced stability and generalization. This cross-domain fusion framework, integrating physics-based mechanism constraints with data-driven fitting, provided a novel pathway for addressing nonlinear transfer challenges in high-precision assembly applications.

As a critical component of aircraft energy systems, the precision of fuel quantity measurement within aircraft fuel tanks is paramount, directly influencing flight safety, operational stability, and mission range planning. However, to optimize the aerodynamic layout and maximize space utilization within the fuselage, modern aircraft fuel tanks are frequently designed with highly irregular geometries. These containers often incorporate complex internal structures, including stiffeners, baffles, and surge boxes, which significantly complicate the relationship between liquid level height and actual fuel volume. Consequently, during dynamic flight maneuvers such as climbing, diving, or rolling, traditional measurement techniques relying solely on single-point liquid-level sensors fail to accurately reflect the true fuel quantity, leading to potential safety hazards and inefficient fuel management. To address the challenging fuel-volume calculation within tanks possessing such complex internal architectures, this research introduced an innovative volume calculation algorithm based on a mesh folding technique. This approach bypassed the need for simplified geometric assumptions, instead leveraging high-fidelity three-dimensional data. The computational procedure was systematic and robust: initially, comprehensive mesh topology information was extracted directly from the fuel tank’s STL (STereoLithography) format file. Subsequently, by integrating real-time aircraft attitude data, specifically pitch and roll angles, with sensor-derived liquid level, the exact spatial position and orientation of the fuel liquid plane were determined. Once the liquid plane was established, the algorithm identified all mesh elements that intersected with this plane. These specific elements underwent a precise division and reconstruction process to align perfectly with the fluid boundary. Following this reconstruction, a unique “folding” operation was executed: all mesh components located above the liquid plane were virtually folded and projected down onto the liquid plane surface. This transformation effectively converted the complex problem of calculating a partial volume within an irregular, tilted container into a straightforward summation of closed mesh volumes. Finally, the volumes of all processed mesh elements below the projected plane were aggregated to yield the precise total fuel volume. When evaluated against conventional volume-calculation strategies, such as the slicing, convex-hull, and voxel methods, the proposed mesh-folding technique demonstrated distinct and significant advantages. The traditional slicing method, although intuitive, was prone to substantial errors during cross-sectional reconstruction when dealing with intricate internal baffles and often requried excessive computational time when high precision was required. The convex hull method was fundamentally limited by its inability to accurately model the non-convex geometric features inherent in most fuel tanks. Similarly, the voxel method faced a persistent dilemma: achieving high resolution requires prohibitive computational resources, whereas lowering resolution compromised accuracy. In contrast, the mesh folding method exhibited exceptional universal applicability. It was not limited by the complexity of internal tank structures and completely eliminated the tedious and time-consuming steps associated with complex cross-sectional mesh reconstruction. As a result, this approach achieved a superior balance, delivering high-precision calculations at speeds suitable for real-time onboard applications. To rigorously validate the performance of this algorithm, extensive comparative simulations were conducted. The proposed method was benchmarked against the slicing, voxel, and Monte Carlo methods across a diverse set of model characteristics, varying both the number of mesh faces and the degree of geometric concavity. These tests confirmed the algorithm’s robustness, demonstrating its consistent accuracy and superior computational efficiency regardless of model complexity. Furthermore, the practical viability of the method was verified through physical experiments on an actual aircraft fuel tank. By simulating various flight attitudes and comparing the algorithm’s output with empirical measurements obtained from controlled filling and weighing tests, the results showed strong consistency between simulation and experiment. This close alignment not only validated the theoretical framework but also demonstrated the method’s readiness for engineering application, offering a reliable technical pathway for the development of next-generation high-precision fuel measurement systems.

This study addressed “social network fatigue” and algorithmic recommendation mechanisms in the context of mobile social media, aiming to reveal the driving mechanism of motion graphics visual design elements on users' willingness to interact on social media. Based on the S-O-R theoretical framework, and drawing on cognitive load theory and Gestalt principles, the study constructed five motion graphics design elements, exploring their impact on user interaction willingness with respect to visual attention and information processing depth. A single-factor, between-subjects questionnaire experiment was conducted, selecting active users from a high-contextuality East Asian culture as subjects, and structural equation modeling was used as the primary method for validation. The study revealed that the visual design of motion graphics exerted an asymmetric impact on interaction willingness. Layout simplicity played a stable supporting role in the continuous process of attention and processing, and was an important condition for forming effective visual order. Visual guidance and graphic relevance enhanced attention and understanding, thereby promoting interaction; animation complexity and information density mainly caused cognitive burden, and while their reduction reduced interference, simply lowering information density alone did not enhance attractiveness. The study further confirmed a chained transformation effect from visual attention to information-processing depth and subsequently to interaction intention, indicating that deep understanding rather than momentary attention was the key driver of user interaction. Based on this mechanism, the study proposed optimization strategies centered on visual-noise reduction and dynamic guidance to enhance the communication effectiveness of social media advertising.