Spectral Features Research Articles

Full-scale semantic segmentation of hyperspectral imaging based on spatial spatial-spectral joint network

Abstract. Hyperspectral images contain dozens or even hundreds of spectral bands, which contain rich spectral information and help distinguish different ground objects. Hyperspectral images have a wide range of applications in urban planning, environmental monitoring, and other fields. The semantic segmentation of hyperspectral images is one of the current research hotspots. The difficulty lies in the rich spectral information and strong correlation of hyperspectral images. Traditional semantic segmentation methods cannot fully extract information, which affects the accuracy of classification. This article utilizes an encoding decoding structure to simultaneously extract deep and shallow features of images. A REGCS convolution module was constructed using the idea of group convolution to extract spectral and spatial features of images. We compared the Salinas Valley dataset and MUUFL dataset with various classification algorithms. The experimental results show that compared with other classification models, the RESSU model has achieved stable and excellent results in hyperspectral image classification experiments. Among them, in the classification experiment of the Salinas Valley dataset, the accuracy of single class classification reached over 92%. In the effectiveness analysis experiment, we calculated different model parameter quantities to verify the performance of our method, and ultimately achieved good results.

A dual-branch selective attention capsule network for classifying kiwifruit soft rot with hyperspectral images

Kiwifruit soft rot is highly contagious and causes serious economic loss. Therefore, early detection and elimination of soft rot are important for postharvest treatment and storage of kiwifruit. This study aims to accurately detect kiwifruit soft rot based on hyperspectral images by using a deep learning approach for image classification. A dual-branch selective attention capsule network (DBSACaps) was proposed to improve the classification accuracy. The network uses two branches to separately extract the spectral and spatial features so as to reduce their mutual interference, followed by fusion of the two features through the attention mechanism. Capsule network was used instead of convolutional neural networks to extract the features and complete the classification. Compared with existing methods, the proposed method exhibited the best classification performance on the kiwifruit soft rot dataset, with an overall accuracy of 97.08% and a 97.83% accuracy for soft rot. Our results confirm that potential soft rot of kiwifruit can be detected using hyperspectral images, which may contribute to the construction of smart agriculture.

A review of the satellite remote sensing techniques for assessment of runoff and sediment in soil erosion

Abstract Soil erosion monitoring is essential for the ecological evaluation and dynamic monitoring of land resources via remote sensing technology. In this paper, we provide new insights into the existing problems and development directions of traditional models, which are supported by new technologies. An important role is played by remote sensing information acquisition technology in the qualitative and quantitative evaluation of soil erosion, and the data and technical support provided are systematically reviewed. We provide a detailed overview of the research progress associated with empirical statistical models and physically driven process models of soil erosion, and the limitations of their application are also summarized. The preliminary integration of remote sensing data sources with high spatial and temporal resolution and new technologies for soil erosion monitoring enables the high-precision quantitative estimation of sediment transport trajectories, the watershed river network density, and the terrain slope, enhancing the accuracy of erosion factor identification, such as spectral feature recognition from erosion information, gully erosion feature extraction, and vegetation coverage estimation. However, the current erosion models, driven by algorithms and models, are not comprehensive enough, particularly in terms of the spatial feature extraction of erosion information, and there are limitations in the applicability and accurate estimation of such models.

Fully Quantum Simulation of Polaritonic Vibrational Spectra of Large Cavity-Molecule System.

The formation of molecular vibrational polaritons, arising from the interplay between molecular vibrations and infrared cavity modes, is a quantum phenomenon necessitating accurate quantum dynamical simulations. Here, we introduce the cavity vibrational self-consistent field/virtual state configuration interaction method, enabling quantum simulation of the vibrational spectra of many-molecule systems within the optical cavity. Focusing on the representative (H2O)21 system, we showcase this parameter-free quantum approach's ability to capture both linear and nonlinear vibrational spectral features. Our findings highlight the growing prominence of molecular couplings among OH stretches and bending excited bands with increased light-matter interaction, revealing distinctive nonlinear spectral features induced by vibrational strong coupling.

Stellar Population Astrophysics (SPA) with the TNG. Measurement of the He I 10830 angstrom line in the open cluster Stock 2

The precise measurement of stellar abundances plays a pivotal role in providing constraints on the chemical evolution of the Galaxy. However, before spectral lines can be employed as reliable abundance indicators, particularly for challenging elements such as helium, they must undergo thorough scrutiny. Galactic open clusters, representing well-defined single stellar populations, offer an ideal setting for unfolding the information stored in the helium spectral line feature. In this study we characterise the profile and strength of the helium transition at around angstrom in nine giant stars in the Galactic open cluster Stock 2. To remove the influence of weak blending lines near the helium feature, we calibrated their oscillator strengths (loggf ) by employing corresponding abundances obtained from simultaneously observed optical spectra. Our observations reveal that the in all the targets is observed in absorption, with line strengths categorised into two groups. Three stars exhibit strong absorption, including a discernible secondary component, while the remaining stars exhibit weaker absorption. The lines are in symmetry and align with or near their rest wavelengths, suggesting a stable upper chromosphere without a significant systematic mass motion. We find a correlation between the strength and the Ca ii logRpHK index, with a slope similar to that reported in previous studies on dwarf stars. This correlation underscores the necessity of accounting for the stellar chromosphere structure when employing as a probe for the stellar helium abundance. The procedure of measuring the we developed in this study is applicable not only to other Galactic open clusters but also to field stars, and we plan to use it to map the helium abundance across various types of stars in the future.

Cation Distribution and Anion Transport in the La3Ga5-xGe1+xO14+0.5x Langasite Structure.

Exploration of compositional disorder using conventional diffraction-based techniques is challenging for systems containing isoelectronic ions possessing similar coherent neutron scattering lengths. Here, we show that a multinuclear solid-state Nuclear Magnetic Resonance (NMR) approach provides compelling insight into the Ga3+/Ge4+ cation distribution and oxygen anion transport in a family of solid electrolytes with langasite structure and La3Ga5-xGe1+xO14+0.5x composition. Ultrahigh field 71Ga Magic Angle Spinning (MAS) NMR experiments acquired at 35.2 T offer striking resolution enhancement, thereby enabling clear detection of Ga sites in different coordination environments. Three-connected GaO4, four-connected GaO4 and GaO6 polyhedra are probed for the parent La3Ga5GeO14 structure, while one additional spectral feature corresponding to the key (Ga,Ge)2O8 structural unit which forms to accommodate the interstitial oxide ions is detected for the Ge4+-doped La3Ga3.5Ge2.5O14.75 phase. The complex spectral line shapes observed in the MAS NMR spectra are reproduced very accurately by the NMR parameters computed for a symmetry-adapted configurational ensemble that comprehensively models site disorder. This approach further reveals a Ga3+/Ge4+ distribution across all Ga/Ge sites that is controlled by a kinetically governed cation diffusion process. Variable temperature 17O MAS NMR experiments up to 700 °C importantly indicate that the presence of interstitial oxide ions triggers chemical exchange between all oxygen sites, thereby enabling atomic-scale understanding of the anion diffusion mechanism underpinning the transport properties of these materials.

High-resolution soil erosion mapping in croplands via Sentinel-2 bare soil imaging and a two-step classification approach

Erosion-induced lateral soil redistribution leads to spatially heterogenous soil composition, which can be captured through the distinctive spectral reflectance of soils under varying levels of erosion influence. This points to the potential of using remotely sensed soil spectra to detect severe erosion and deposition hotspots in exposed croplands and, importantly, further differentiate the intra-class spectral variability of moderate erosion that often occupies the largest proportion. Here, we aim to develop a two-step erosion classification and mapping approach based on multitemporal compositing of Sentinel-2 bare soil images of a typical agricultural region (11,500 km2) of northeast China. A random forest classifier was firstly trained against the ground-truth data derived from very high resolution (VHR) imagery in Google Earth, with an overall accuracy of 91 % that allowed for clear delineation of severe erosion and deposition areas based on their distinct topographic and spectral features particularly in the red and red-edge bands. In the second step, the remaining area of moderate erosion (60.30 %) was further differentiated using Iterative Self-Organizing cluster unsupervised classification to yield a five-class soil erosion map at 10 m spatial resolution. The accuracy of the predicted map was successfully validated by independent Caesium-137 (137Cs) and soil organic carbon observations at catchment and regional scales, as revealed by significant inter-class differences in soil redistribution rates estimated from 137Cs inventory. The severe erosion class had a soil loss rate of 5.5 mm yr−1, suggesting that previous assessments have underestimated erosion severity. The spatial accordance of crop growth with soil erosion intensity, particularly in localized settings, further highlighted the potential of bare soil imaging for mapping the spatiotemporal development of soil erosion and its response to targeted sustainable cropland management efforts.

Opportunities and Challenges: Process Raman for the Real-Time Release Testing (RTRT) of Extended-Release Formulations

The use of a process analytical technology has been demonstrated using near-infrared spectroscopy for continuous manufacturing of pharmaceutical formulations and is within the scope of the U.S. Food and Drug Administration's real-time release testing initiative. While effective for simple formulations, this preliminary study investigates whether such a spectroscopic surrogate application can replace pharmaceutical dissolution testing for extended-release formulations. In this study, we will assess the use of process Raman spectroscopy for real-time dissolution testing. Extended-release tablet formulations often accomplish the release rate delay through the addition of gelling agents. In this work, hydroxypropyl methylcellulose (HPMC) polymers were used to formulate extended-release niacin tablets. Process Raman spectroscopy was evaluated as a tool to effectively model dissolution profiles to determine if the optical technique has the ability to differentiate HPMC polymers from the background and be selective for the polymer type employed. Our preliminary work indicates that while Raman can effectively detect and monitor the niacin response of the tablet formulations, there are not enough unique spectral features between the different HPMC polymers to selectively resolve their responses. Additional measurements and chemometric analysis might suggest otherwise. Thus, for extended-release tablet applications with continuous manufacturing, further dissolution surrogate development is needed.

Low-donor content approach to achieve sensitive narrowband organic photodetectors

Narrowband near-infrared (NIR) detectors are key components for various modern biomedical, agricultural, astronomical sensing as well as monitoring applications. Typically, a narrowband organic photodetector (OPD) requires thick junctions or additional optical elements, which are either tedious or cost-ineffective. Here, we adopt the low-donor content strategy in our thin-film device to realize narrowband spectral features. The much-mitigated aggregation drastically improves shunt resistance and dark current while maintaining decent photocurrent. Upon further dilution, a thinner electron transporting layer can effectively shorten the transient response to reach fast speed. n-i-p and p-i-n device configurations are also compared, which exhibits drastic changes for low-donor content OPDs. The 5 wt% D:A thin-film bulk-heterojunction OPD presents an external quantum efficiency of 49.3% at 770 nm with zero bias, and a high thermal-noise calculated specific detectivity (D*) of 1.0 × 1013 Jones (7.1 × 1012 Jones based on noise spectral density) at 0 V. Such narrowband OPD also shows a fast NIR response of 530 ns/840 ns (rise/fall) when operated at 10 kHz. The results reveal the potential for low-donor-content blend in realizing narrowband photodetection.

Effects of phantom microstructure on their optical properties.

Developing stable, robust, and affordable tissue-mimicking phantoms is a prerequisite for any new clinical application within biomedical optics. To this end, a thorough understanding of the phantom structure and optical properties is paramount. We characterized the structural and optical properties of PlatSil SiliGlass phantoms using experimental and numerical approaches to examine the effects of phantom microstructure on their overall optical properties. We employed scanning electron microscope (SEM), hyperspectral imaging (HSI), and spectroscopy in combination with Mie theory modeling and inverse Monte Carlo to investigate the relationship between phantom constituent and overall phantom optical properties. SEM revealed that microspheres had a broad range of sizes with average and were also aggregated, which may affect overall optical properties and warrants careful preparation to minimize these effects. Spectroscopy was used to measure pigment and SiliGlass absorption coefficient in the VIS-NIR range. Size distribution was used to calculate scattering coefficients and observe the impact of phantom microstructure on scattering properties. The results were surmised in an inverse problem solution that enabled absolute determination of component volume fractions that agree with values obtained during preparation and explained experimentally observed spectral features. HSI microscopy revealed pronounced single-scattering effects that agree with single-scattering events. We show that knowledge of phantom microstructure enables absolute measurements of phantom constitution without prior calibration. Further, we show a connection across different length scales where knowledge of precise phantom component constitution can help understand macroscopically observable optical properties.

Stable Isotope Probing-nanoFTIR for Quantitation of Cellular Metabolism and Observation of Growth-Dependent Spectral Features.

This study utilizes nanoscale Fourier transform infrared spectroscopy (nanoFTIR) to perform stable isotope probing (SIP) on individual bacteria cells cultured in the presence of 13C-labelled glucose. SIP-nanoFTIR simultaneously quantifies single-cell metabolism through infrared spectroscopy and acquires cellular morphological information via atomic force microscopy. The redshift of the amide I peak corresponds to the isotopic enrichment of newly synthesized proteins. These observations of single-cell translational activity are comparable to those of conventional methods, examining bulk cell numbers. Observing cells cultured under conditions of limited carbon, SIP- nanoFTIR is used to identify environmentally-induced changes in metabolic heterogeneity and cellular morphology. Individuals outcompeting their neighboring cells will likely play a disproportionately large role in shaping population dynamics during adverse conditions or environmental fluctuations. Additionally, SIP-nanoFTIR enables the spectroscopic differentiation of specific cellular growth phases. During cellular replication, subcellular isotope distribution becomes more homogenous, which is reflected in the spectroscopic features dependent on the extent of 13C-13C mode coupling or to specific isotopic symmetries within protein secondary structures. As SIP-nanoFTIR captures single-cell metabolism, environmentally-induced cellular processes, and subcellular isotope localization, this technique offers widespread applications across a variety of disciplines including microbial ecology, biophysics, biopharmaceuticals, medicinal science, and cancer research.

Understanding Spin‐Triplet Excited States in Carbene‐Metal‐Amides

Carbene‐metal‐amides (CMAs) are emerging delayed fluorescence materials for organic light‐emitting diode (OLED) applications. CMAs possess fast, efficient emission owing to rapid forward and reverse intersystem crossing (ISC) rates. The resulting dynamic equilibrium between singlet and triplet spin manifolds distinguishes CMAs from most purely organic thermally activated delayed fluorescence emitters. However, direct experimental triplet characterization in CMAs is underutilized, limiting our detailed understanding of the ISC mechanism. In this work, we combine time‐resolved spectroscopy with tuning of state energies through environmental polarity and metal substitution, focusing on the interplay between charge‐transfer (3CT) and local exciton (3LE) triplets. Unlike previous photophysical work, we investigate evaporated host:guest films of CMAs and small‐molecule hosts for increased device relevance. Transient absorption reveals an evolution in the triplet excited‐state absorption (ESA) consistent with a change in orbital character between hosts with differing dielectric constants. Using quantum chemical calculations, we simulate ESAs of the lowest triplet states, highlighting the contribution of only 3CT and donor‐moiety 3LE states to spectral features, with no strong evidence for a low‐lying acceptor‐centered 3LE. Thus, our work provides a blueprint for understanding the role of triplet excited states in CMAs which will enable further intelligent optimization of this promising class of materials.

Understanding Spin-Triplet Excited States in Carbene-Metal-Amides.

Carbene-metal-amides (CMAs) are emerging delayed fluorescence materials for organic light-emitting diode (OLED) applications. CMAs possess fast, efficient emission owing to rapid forward and reverse intersystem crossing (ISC) rates. The resulting dynamic equilibrium between singlet and triplet spin manifolds distinguishes CMAs from most purely organic thermally activated delayed fluorescence emitters. However, direct experimental triplet characterization in CMAs is underutilized, limiting our detailed understanding of the ISC mechanism. In this work, we combine time-resolved spectroscopy with tuning of state energies through environmental polarity and metal substitution, focusing on the interplay between charge-transfer (3CT) and local exciton (3LE) triplets. Unlike previous photophysical work, we investigate evaporated host:guest films of CMAs and small-molecule hosts for increased device relevance. Transient absorption reveals an evolution in the triplet excited-state absorption (ESA) consistent with a change in orbital character between hosts with differing dielectric constants. Using quantum chemical calculations, we simulate ESAs of the lowest triplet states, highlighting the contribution of only 3CT and donor-moiety 3LE states to spectral features, with no strong evidence for a low-lying acceptor-centered 3LE. Thus, our work provides a blueprint for understanding the role of triplet excited states in CMAs which will enable further intelligent optimization of this promising class of materials.

Automatic detection method of bladder tumor cells based on color and shape features

Bladder urothelial carcinoma is the most common malignant tumor disease in urinary system, and its incidence rate ranks ninth in the world. In recent years, the continuous development of hyperspectral imaging technology has provided a new tool for the auxiliary diagnosis of bladder cancer. In this study, based on microscopic hyperspectral data, an automatic detection algorithm of bladder tumor cells combining color features and shape features is proposed. Support vector machine (SVM) is used to build classification models and compare the classification performance of spectral feature, spectral and shape fusion feature, and the fusion feature proposed in this paper on the same classifier. The results show that the sensitivity, specificity, and accuracy of our classification algorithm based on shape and color fusion features are 0.952, 0.897, and 0.920, respectively, which are better than the classification algorithm only using spectral features. Therefore, this study can effectively extract the cell features of bladder urothelial carcinoma smear, thus achieving automatic, real-time, and noninvasive detection of bladder tumor cells, and then helping doctors improve the efficiency of pathological diagnosis of bladder urothelial cancer, and providing a reliable basis for doctors to choose treatment plans and judge the prognosis of the disease.

Early Detection of Rubber Tree Powdery Mildew by Combining Spectral and Physicochemical Parameter Features

Powdery mildew significantly impacts the yield of natural rubber by being one of the predominant diseases that affect rubber trees. Accurate, non-destructive recognition of powdery mildew in the early stage is essential for the cultivation management of rubber trees. The objective of this study is to establish a technique for the early detection of powdery mildew in rubber trees by combining spectral and physicochemical parameter features. At three field experiment sites and in the laboratory, a spectroradiometer and a hand-held optical leaf-clip meter were utilized, respectively, to measure the hyperspectral reflectance data (350–2500 nm) and physicochemical parameter data of both healthy and early-stage powdery-mildew-infected leaves. Initially, vegetation indices were extracted from hyperspectral reflectance data, and wavelet energy coefficients were obtained through continuous wavelet transform (CWT). Subsequently, significant vegetation indices (VIs) were selected using the ReliefF algorithm, and the optimal wavelengths (OWs) were chosen via competitive adaptive reweighted sampling. Principal component analysis was used for the dimensionality reduction of significant wavelet energy coefficients, resulting in wavelet features (WFs). To evaluate the detection capability of the aforementioned features, the three spectral features extracted above, along with their combinations with physicochemical parameter features (PFs) (VIs + PFs, OWs + PFs, WFs + PFs), were used to construct six classes of features. In turn, these features were input into support vector machine (SVM), random forest (RF), and logistic regression (LR), respectively, to build early detection models for powdery mildew in rubber trees. The results revealed that models based on WFs perform well, markedly outperforming those constructed using VIs and OWs as inputs. Moreover, models incorporating combined features surpass those relying on single features, with an overall accuracy (OA) improvement of over 1.9% and an increase in F1-Score of over 0.012. The model that combines WFs and PFs shows superior performance over all the other models, achieving OAs of 94.3%, 90.6%, and 93.4%, and F1-Scores of 0.952, 0.917, and 0.941 on SVM, RF, and LR, respectively. Compared to using WFs alone, the OAs improved by 1.9%, 2.8%, and 1.9%, and the F1-Scores increased by 0.017, 0.017, and 0.016, respectively. This study showcases the viability of early detection of powdery mildew in rubber trees.

In-situ determination of micro-hardness in laser cladding of stellite using optical emission spectroscopy

Offline measurements in Laser Cladding (LC) processes, such as micro-hardness measurement, are destructive and time-consuming tests. Therefore, in-situ monitoring of clads parameters is of significant importance. In this study, we investigate the plasma plume emission produced during the LC process of Stellite 6 powder on a low-carbon steel substrate and its correlation with the micro-hardness of the clads over a wide range of LC parameters. It is demonstrated that a strong correlation between the spectral features in the UV region of 280–320 nm and micro-hardness exists, so that, the micro-hardness of samples can be determined by the total intensity of light in this range, within an average tolerance of 10 %. An exponential relation between hardness and UV radiation is empirically obtained that works very well for a wide range of dilutions, from 20 % to 70 %. It means that the UV spectra even determine the micro-hardness of samples with dilutions above 60 %, where the process shifts towards laser powder welding. This radiation is suggested to originate from the collisions of electrons with nano-sized clusters, which are produced during the laser interaction with the melt pool. The produced nanoparticles are the source of UV radiation and scatters the driving laser light, at the same time, processes that qualitatively explain the exponential relationship between the hardness and the UV light intensity.

CB-HDM: ECG signal based heart disease classification using convolutional block attention assisted hybrid deep Maxout network

Cardiovascular diseases (CVD) stand as the leading cause of death globally. Diagnosis of CVD relies on electrocardiograms (ECGs), which record the heart’s activity. Adequate labeled data is essential for effectively training machine learning techniques, which may also demand considerable time and memory resources. To address these existing issues, this research introduces an improved detection technique for predicting CVD. Initially, data is sourced from the MIT-BIH arrhythmia database, followed by pre-processing using the Modified FIR filter model (M-FIR) to filter ECG signals. It is used to enhance signal quality and eliminate unwanted noise. M-FIR is used to preserve the specific features of the signal. Feature extraction utilizes the Radial Hilbert Function Transform Network (RHFTN) technique. It can extract the temporal and spectral signal features for effective CVD detection. The Improved Bald Eagle Search Algorithm (IBES) is employed to select features and minimize irrelevant features. Detection and classification of heart disease from ECG signals are achieved using the Convolutional Block Attention Assisted Hybrid Deep Maxout Network model (CB-HDM). Losses in the network model are mitigated by the Gazelle Optimization Algorithm (GOA). The evaluation results are simulated using Python. The proposed model achieved 98.81% accuracy, 96.18% precision, 96.87% recall, and a 94.39% F1 score. The proposed model demonstrates superior efficiency compared to other classifiers.

Identification of rare earth elements in synthetic and natural monazite and xenotime by visible-to-shortwave infrared reflectance spectroscopy

To support the role of proximal and remote sensing in geological rare earth element (REE) resource exploration, we studied the reflectance spectroscopy of synthetic single- and mixed-REE phosphate phases. Synthesis yielded monazite for the elements La to Gd, and xenotime for Dy to Lu and Y. Visible-to-shortwave infrared (350–2500 nm) reflectance spectra of synthetic single-REE monazites and xenotimes can be used to identify the ions responsible for the absorption features in natural monazites and xenotimes. Nd3+, Pr3+ and Sm3+ produce the main absorption features in monazites. In natural xenotime, Dy3+, Er3+, Ho3+ and Tb3+ ions cause the prevalent absorptions. The majority of the REE-related absorption features are due to photons exciting electrons within the 4f subshell of the trivalent lanthanide ions to elevated energy levels resulting from spin-orbit coupling. There are small (< 20 nm) shifts in the wavelengths of these absorptions depending on the nature of the ligands. The energy levels are further split by crystal field effects, manifested in the reflectance spectra as closely spaced (∼ 5–20 nm) multiplets within the larger absorption features. Superimposed on the electronic absorptions are vibrational absorptions in the H2O molecule or within [OH]−, [CO3]2− and [PO4]3− functional groups, but so far only the carbonate-related spectral features seem usable as a diagnostic tool in REE-bearing minerals. Altogether, our study creates a strengthened knowledge base for detection of REE using reflectance spectroscopy and provides a starting point for the identification of REE and their host minerals in mineral resources by means of hyperspectral methods.

Infrared spectral features of the charge carriers in nickel salen polymeric complexes

A series of the polymeric complexes (PCs) of nickel(II) with N2O2 Schiff base ligands have been studied by in situ Fourier Transform Infrared (FTIR) spectroelectrochemistry. The molecular structure of the metal-salen complexes as starting materials for the polymer synthesis was modulated by various substituents in the phenolate moieties and diimine bridge. Two groups of the infrared-active vibration (IRAV) bands are obtained in the spectra depending on the oxidation level of the polymers. The assignment of the main IRAV bands was carried out and the electrochemically generated charge carriers were determined. Analysis of the spectroscopic features observed in the mid and near IR region provides information on the charge distribution within the polymer film.

Source Tracing of Kidney Injury via the Multispectral Fingerprint Identified by Machine Learning-Driven Surface-Enhanced Raman Spectroscopic Analysis.

Early diagnosis of drug-induced kidney injury (DIKI) is essential for clinical treatment and intervention. However, developing a reliable method to trace kidney injury origins through retrospective studies remains a challenge. In this study, we designed ordered fried-bun-shaped Au nanocone arrays (FBS NCAs) to create microarray chips as a surface-enhanced Raman scattering (SERS) analysis platform. Subsequently, the principal component analysis (PCA)-two-layer nearest neighbor (TLNN) model was constructed to identify and analyze the SERS spectra of exosomes from renal injury induced by cisplatin and gentamycin. The established PCA-TLNN model successfully differentiated the SERS spectra of exosomes from renal injury at different stages and causes, capturing the most significant spectral features for distinguishing these variations. For the SERS spectra of exosomes from renal injury at different induction times, the accuracy of PCA-TLNN reached 97.8% (cisplatin) and 93.3% (gentamicin). For the SERS spectra of exosomes from renal injury caused by different agents, the accuracy of PCA-TLNN reached 100% (7 days) and 96.7% (14 days). This study demonstrates that the combination of label-free exosome SERS and machine learning could serve as an innovative strategy for medical diagnosis and therapeutic intervention.