To address the enhanced matrix effects and intensified nonlinear spectral responses caused by the difficulty of grinding large-particle coal in industrial settings,this study proposes a hybrid PLS-AE-RR predictive model based on the fusion of near-infrared spectroscopy(NIRS)and X-ray fluorescence(XRF)spectra,aimed at improving the accuracy of on-line calorific-value analysis.The method implements a three-stage hybrid framework--linear baseline+nonlinear feature extraction+residual correction--where partial least squares regression(PLS)first models the global linear relationship between the fused spectra and calorific value;an autoencoder(AE)then extracts low-dimensional nonlinear representations that PLS cannot capture;and finally ridge regression(RR)fits and corrects the nonlinear residuals.Exper-imental validation using 153 blended coal samples from power plants demonstrates breakthrough perfor-mance in calorific value prediction for large-particle-size coal.On the test set,determination coefficients(R²)for lignite and bituminous coal reached 0.974 and 0.938,respectively,with mean absolute errors of 0.233 MJ/kg and 0.216 MJ/kg.The proposed method significantly outperforms standalone PLS and al-ternative nonlinear correction models,confirming the generalization advantage of ridge regression in residu-al fitting.Consequently,this achievement provides a grinding-free,high-precision online analysis solution for raw coal calorific value in coal-fired power plants,offering critical technical support for refined fuel man-agement and operational optimization.
The workflow begins with approximate orthorectification of aerial images using coarse position-ing results and flight parameters,followed by extraction of the corresponding satellite image regions.RepVGG is employed to extract coarse image features,and initial correspondences are established via nearest-neighbor matching.Candidate pairs are then filtered using MiHo clustering and normalized cross-correlation(NCC).The refined correspondences are further enhanced by a Transformer module to achieve precise alignment.Based on the resulting high-precision matches,an angular error correction ma-trix is constructed,and iterative refinement is applied to effectively compensate for systematic errors.Ex-perimental results indicate that the proposed method substantially improves positioning accuracy relative to conventional approaches,achieving an approximately 70%enhancement in representative scenarios;even at a slant range of 90 km,the positioning error remains approximately 120 m.By leveraging image match-ing under challenging viewing geometries,the proposed passive localization method provides a novel solu-tion to limitations of conventional algorithms.A two-step matching strategy integrating Transformer-based feature enhancement with MiHo clustering-based filtering is introduced,effectively mitigating the ac-curacy degradation of traditional passive localization under high-altitude,large-oblique viewing conditions in the presence of small angular errors.
A commercially available transparent phone case was used as the substrate,and an Ag@Cu mi-cro-/nano-composite antibacterial coating was fabricated on its surface via laser-induced backward transfer(LIBT).The results indicated a positive correlation between laser power and the elemental content of the microstructured coating,with the mass ratios of Ag and Cu increasing as laser power increased.This ap-proach enables efficient fabrication under ambient atmospheric conditions with a simple and cost-effective process.The resulting microstructured coating exhibited pronounced antibacterial activity against both Gram-positive Staphylococcus aureus and Gram-negative Escherichia coli.At laser powers of≥16 W,the antibacterial rate exceeded 99.99%,and the increased copper mass ratio further enhanced antibacterial efficacy.These findings address key limitations of conventional technologies by enabling high-precision construction of micro-/nano-structures on transparent substrates.Moreover,the processing parameters are precisely controllable,supporting scalable manufacturing and offering a general strategy for functional-izing the surfaces of other transparent organic materials.Overall,a practical route for producing antibacte-rial phone cases is established,providing methodological guidance for the design and fabrication of ad-vanced antibacterial surfaces and addressing bacterial contamination and insufficient antibacterial perfor-mance in mobile phone casings.
To improve the 3D visualization effect and autostereoscopic display performance of medical im-ages,this paper proposed a light field display method for medical images based on virtual light field acquisi-tion.First,based on medical tomographic image data,preprocessing algorithms were adopted to process the original image data.Combined with surface reconstruction and the improved Quadric Error Metrics(QEM)simplification algorithm,a lightweight 3D medical model was constructed while its key structural features were preserved.Then,aiming at the limitations of traditional light field acquisition methods,a virtual camera array construction strategy based on computer graphics was further designed.It generated high-quality virtual light field content without relying on physical optical equipment.Finally,the light field acquisition scene was built for the simplified 3D model using multi-view rendering technology.It opti-mized the acquisition efficiency and spatial resolution of light field images.Experimental results show that the rendering overhead of the medical model is reduced by 90.16%.And the geometric accuracy of the model is only lost by 4.06%.The virtual light field acquisition system achieved rendering efficiencies of 9.83 frames per second(FPS)and 9.57 FPS on the two optical experimental platforms,respectively,with a substantial enhancement in the generation efficiency of elemental images,thereby largely fulfilling the real-time acquisition criteria.The proposed method significantly reduces the model complexity while ensuring the geometric accuracy of the medical model,and effectively improves the quality and real-time performance of virtual light field display.
To address the issues of complex crack morphologies,environmental interference,and the im-balance between detection accuracy and model lightweight requirements in road surface inspection,this pa-per proposed a lightweight road crack detection method with adaptive feature extraction.First,a Crack Ef-ficient Attention(CEA)module was designed based on the slender shape and large span characteristics of cracks,compressing feature dimensions to capture long-distance spatial dependencies.Second,a Dynamic Sampling Feature Pyramid Network(DSFPN)was constructed for adaptive sampling and target feature extraction,enhancing representation capability for heterogeneous crack features.Third,the HGNet_GS lightweight backbone network was improved,and a CEA Group Head(CGHead)was proposed,signifi-cantly reducing computational redundancy;the PIoU(Powerful IoU)loss function was adopted to solve anchor box expansion problems and improve convergence speed for small models.Additionally,a civilian road defect dataset containing 2 985 images under various lighting conditions was established to validate model generalization.Experimental results show that compared with the baseline YOLOv8n model,the proposed method reduces parameters and computational cost by 50%and 52%,respectively;on the self-built dataset,mAP50 and mAP95 increase by 5.4%and 4.1%;on the public RDD2022 dataset,these metrics improve by 2.1%and 1.5%.The model has been deployed on edge devices and verified through engineering tests,demonstrating its capability to meet practical requirements for lightweight road crack de-tection and providing a technical solution for automated road maintenance systems.
