Spectroscopic Techniques in Food Adulteration Detection: A Comprehensive Review

Abstract

Food adulteration, as a highly concealed form of economic fraud, severely endangers food safety, disrupts market order, and infringes on consumer rights and interests. Conventional detection methods are constrained by cumbersome operations, time-consuming procedures and sample destruction, which fail to meet the demands of on-site rapid screening and real-time online monitoring. Benefiting from distinct advantages including rapid detection, non-destructive measurement and environmental friendliness, spectroscopic techniques have exhibited significant application potential and research value in the field of food adulteration detection. Centering on high-value commodities such as meat, edible oil, honey, dairy products and condiments, this paper systematically reviews the research progress of near-infrared spectroscopy, Raman spectroscopy, hyperspectral imaging and other spectroscopic technologies in food adulteration detection. The performances, merits and limitations of various spectroscopic techniques in detecting adulteration of different food matrices are comprehensively analyzed. Furthermore, the future development trends of spectroscopic techniques in this field are prospected. This review provides a systematic technical reference for subsequent academic research and practical application of spectroscopic methods in food adulteration detection.

Keywords:
spectroscopic techniques; food adulteration; rapid detection; non-destructive analysis

1. Introduction

Food safety represents a paramount issue closely associated with national livelihood and social stability. Within the globalized food supply chain, food adulteration, a highly concealed type of economic fraud, has emerged as a prominent hazard that compromises food safety, distorts market order, and violates consumer legitimate rights and interests.¹ In this context, there is an urgent imperative to develop rapid, accurate and efficient detection methodologies for the effective identification and continuous monitoring of food adulteration events, which is critical to safeguarding food authenticity and safety, maintaining fair market competition, and improving regulatory efficiency. Conventional analytical approaches for food adulteration detection, including sensory evaluation,² chemical assay,³ and chromatography-mass spectrometry,⁴ deliver satisfactory detection accuracy. Nevertheless, these methods suffer from several inherent limitations, including tedious operational procedures, time-consuming assays, complicated sample pretreatment, heavy reliance on large laboratory equipment, excessive consumption of chemical reagents, and irreversible sample destruction.²-⁴ Such inherent deficiencies severely restrict their applicability to large-scale and high-throughput on-site screening as well as real-time online monitoring. Furthermore, they fail to accommodate the instantaneity, field applicability and non-destructive testing requirements of modern food quality control, thereby creating an urgent research gap for innovative detection alternatives.

Spectroscopic techniques have gradually evolved into a prevailing research hotspot in food adulteration detection by virtue of their inherent merits of rapid response, non destructive measurement and environmental sustainability.¹ To map the research landscape of this field, a keyword cloud map was generated (Figure 1). As illustrated, terms such as spectroscopy, meat, and honey feature prominently, indicating that the research focus is shifting from traditional linear modeling toward intelligent approaches and on-site detection. These technologies can acquire spectral fingerprints of intact samples and realize efficient adulteration identification via chemometric data analysis without causing any sample damage. Additionally, spectroscopic methods feature simple operating protocols, minimal pretreatment requirements, and great feasibility for real-time online monitoring, offering an innovative alternative for routine food quality control. With the combined advantages of facile operation, low cost, and outstanding practicability for field rapid screening and in situ monitoring, spectroscopic techniques effectively compensate for the bottlenecks of conventional detection strategies. This endows them with broad application prospects in modern food industries and equips regulatory authorities with a high-efficiency technical tool to address the increasingly complex challenges in food safety governance.⁵

Figure 1
Keyword cloud map of research on the application of spectroscopic techniques in food adulteration detection.

Herein, this review systematically summarizes recent applications, research advances and core data processing algorithms of mainstream spectroscopic techniques, including near-infrared spectroscopy, hyperspectral imaging and Raman spectroscopy, in the adulteration detection of typical high-value food commodities covering meat, edible oil, honey, dairy products and condiments. By comprehensively comparing the strengths and inherent limitations of different spectroscopic modalities, we elaborate the application potential and cutting-edge progress of spectroscopic detection in this field. This work aims to provide systematic references and insightful perspectives for subsequent fundamental research and practical deployment, and further facilitate the translational progression of spectroscopic technologies from laboratory exploration to large-scale and reliable industrial implementation.

2. Fundamental Principles of Spectroscopic Techniques

2.1. Near-infrared spectroscopy

Near-infrared (NIR) spectroscopy is based on the absorption of NIR radiation by overtone and combination vibrations of chemical bonds in molecular structures, primarily hydrogen-containing groups such as C-H, O-H, and N-H.⁶ When a sample is irradiated with NIR radiation, light at specific wavelengths is selectively absorbed by chemical bonds, generating characteristic absorption spectra that reflect the molecular structure and content of organic components within the sample. Featuring non-destructive measurement and high-throughput capability, NIR spectroscopy is particularly suitable for online monitoring and large-scale screening in the food industry.⁷

2.2. Hyperspectral imaging technology

Hyperspectral imaging integrates spectral analysis and image processing techniques. It scans samples across dozens to hundreds of continuous, narrow spectral bands and records a complete spectral curve for each spatial pixel, simultaneously acquiring spatial and spectral information to construct three-dimensional data cubes.⁸ Unlike conventional spectroscopy, which only analyzes chemical composition, this technique further enables intuitive visualization of the spatial distribution of various components within samples. In food detection, it is uniquely applicable to identifying local adulteration in heterogeneous food matrices and can achieve visualized localization of adulterated regions via pseudo-color imaging.⁹

2.3. Raman spectroscopy

Raman spectroscopy is based on the Raman scattering effect. When a sample is irradiated with a monochromatic laser, most photons undergo elastic scattering, while approximately one in a million photons collides inelastically with sample molecules and exchanges energy, resulting in altered scattered light frequencies, defined as Raman scattering.¹⁰ Such frequency shifts are directly correlated with intrinsic molecular vibrational energy levels, providing specific information regarding molecular structure, chemical bond types, and crystal morphology. Complementary to infrared spectroscopy, Raman spectroscopy is insensitive to moisture and requires minimal sample pretreatment. It is highly applicable to the analysis of water-rich food matrices and enables effective identification of adulterants with similar chemical structures.¹¹

2.4. Mid-infrared spectroscopy

Mid-infrared (MIR) spectroscopy originates from the fundamental vibrational absorption of MIR light by molecular chemical bonds. Absorption peaks emerge when the frequency of incident infrared light matches the intrinsic vibration frequency of chemical bonds. Distinct chemical bonds and functional groups exhibit unique absorption frequencies and intensities; thus, MIR spectra directly characterize the functional group composition and molecular structure of substances. The application of Fourier transform infrared (FTIR) spectrometers has substantially improved the signal-to-noise ratio and scanning speed of this technique. In food research, MIR spectroscopy is widely employed for adulterant identification,¹² major component quantification,¹³ and geographical origin tracing.¹⁴

3. Applications of Spectroscopic Techniques in the Detection of Food Adulteration

In recent years, detection techniques integrating spectroscopic technology with chemometrics and machine learning algorithms have emerged as the core approach for food adulteration detection. These methods offer advantages such as rapid analysis, non-destructive testing, and minimal sample preparation, effectively addressing the limitations of conventional methods, such as chromatography and mass spectrometry, which are often time-consuming and costly. Such approaches have been successfully applied to the detection of adulteration in meat, honey, dairy products, edible oils, and other food categories. The operating principle of food adulteration detection is illustrated in Figure 2. Initially, adulterated and authentic samples are collected, and spectral data are acquired using instruments such as infrared spectroscopy, Raman spectroscopy, and hyperspectral imaging (the acquisition process is shown in Figure 3). The raw spectral data are then smoothed using preprocessing techniques like MSC (multiplicative scatter correction) and SNV (standard normal variate transformation). Subsequently, machine learning models such as SVM (support vector machine) and KNN (k-nearest neighbors) receive the preprocessed data, learn discriminative features, and generate prediction outcomes, thereby achieving effective food adulteration detection. This section reviews the applications of spectroscopic techniques in various food sectors, including meat, edible oils, honey, and dairy products.

Figure 2
Schematic diagram of the principle of food adulteration detection based on spectroscopic techniques.

Figure 3
Spectral data acquisition and processing workflow.

3.1. Detection of meat adulteration

Meat adulteration remains a prevalent fraudulent issue in the global food supply chain, severely undermining consumer rights and posing potential public health risks. The development of rapid, accurate, and non-destructive detection technologies is therefore critical to guaranteeing the authenticity and quality of meat products. In recent years, spectroscopic techniques, especially NIR spectroscopy, hyperspectral imaging, and Raman spectroscopy, have demonstrated enormous application potential due to their inherent advantages of speed, non-destructive nature, high efficiency, and environmental compatibility. Table 1 summarizes selected representative studies.

Table 1
Comparison of reported studies on the application of spectroscopy techniques for meat adulteration detection

Detection object	Detection technology	Analytical performance	Reference
Mutton adulterated with heterogeneous meat	NIR	accuracy rate in the validation set reaches 100%	15
Minced beef products adulterated with oat flour and corn starch	NIR hyperspectral imaging	fast and non-destructive detection is achieved	16
Beef adulteration	HSI	suitable for high-throughput screening	17
Mutton adulterated with fox meat	HSI	R2 = 0.928, RMSEP = 3.00%, RPD = 4.85	18
Beef mince adulterated with chicken, duck, and pork	HSI	prediction set: R2 = 0.98 and RMSEP = 1.8651 for chicken, duck, and pork adulteration	19
Fresh beef adulterated with clenbuterol (salbutamol)	Raman spectroscopy	limit of detection 0.01 mg kg-1, R2 = 0.912	20
Pure meat mince and mixed meat mince (pig, cattle, sheep)	Raman spectroscopy	high classification accuracy	21
Complete red meat samples (90 samples)	Raman spectroscopy	SVM sensitivity > 87% (linear) and > 90% (non-linear); specificity > 88%; PLS-DA classification accuracy > 80%	22

NIR spectroscopy discriminates meat adulterants and quantifies compositions by capturing spectral variations derived from overtone and combination absorption of hydrogen-bearing functional groups. Zuo et al.¹⁵ collected spectral data of 360 mutton samples in the visible-NIR (350 1000 nm) and NIR (1000-1700 nm) regions. By combining first derivative correction with multiplicative scatter correction, they established a qualitative discrimination model for foreign meat adulteration in mutton, achieving 100% accuracy in the validation set. Oliveira et al.¹⁶ evaluated a NIR hyperspectral imaging system coupled with chemometrics for non-destructive and rapid detection of meat thickeners (oat flour and corn starch) in minced beef. Partial least squares regression (PLSR) models were constructed using full-spectrum data and key wavelength subsets screened via variable importance in projection (VIP) and competitive adaptive reweighted sampling (CARS).

Hyperspectral imaging integrates conventional spectral analysis with computer vision, acquiring two-dimensional spatial images and one-dimensional spectral information over continuous bands to realize qualitative identification, quantitative determination, and spatial distribution visualization of adulterants. Gao et al.¹⁷ proposed a novel non-destructive method combining hyperspectral imaging with independent component analysis to extract tissue surface features for beef adulteration detection, showing great industrial prospects for high-throughput screening in meat processing and quality assurance operations. Bai et al.¹⁸ selected 14 characteristic wavelengths using hyperspectral technology coupled with two-dimensional correlation spectroscopy (2D-COS). The established support vector regression (SVR) model achieved a coefficient of determination (R²) of 0.928, a root mean square error of prediction (RMSEP) of 3.00%, and a residual predictive deviation (RPD) of 4.85 for detecting fox meat adulteration in mutton, enabling precise quantitative detection. Li et al.¹⁹ adopted a fusion strategy of hyperspectral reflectance and transmittance to detect adulteration of minced beef with chicken, duck, and pork. Combined with PLSR via intermediate-level data fusion, R² values reached 0.9845, 0.9860, and 0.9751, with RMSEP values as low as 1.8651, 1.7711, and 2.3665, respectively, significantly outperforming single spectral models.

Raman spectroscopy generates characteristic molecular fingerprint spectra by analyzing molecular vibrational and rotational information based on inelastic light scattering, supporting qualitative discrimination and quantitative analysis of adulterants. Zhai et al.²⁰ employed surface-enhanced Raman spectroscopy (SERS) for rapid detection of clenbuterol (salbutamol) in fresh beef, achieving a limit of detection as low as 0.01 mg kg^-1 and a correlation coefficient of 0.912, satisfying the requirements for trace adulterant detection. Nedeljkovic et al.²¹ combined Raman spectroscopy with machine learning to accurately distinguish pure and mixed meat batters. Raman spectra were acquired from 19 sample groups with different blending ratios of pork, beef, and mutton, and sample homogenization was proven to significantly improve spectral consistency and classification accuracy. Robert et al.²² analyzed 90 intact red meat samples using Raman spectroscopy. Spectral preprocessing was performed using rubber band baseline correction, Savitzky-Golay smoothing, and standard normal variate transformation. SVM models with linear and nonlinear kernels yielded sensitivity above 87 and 90%, respectively, with corresponding specificity exceeding 88% in the test set; the partial least squares discriminant analysis (PLS-DA) model achieved a classification accuracy over 80% for individual meat categories.

Spectroscopic techniques exhibit prominent superiority in meat adulteration detection, enabling rapid, non-destructive, and online detection with minimal sample pretreatment, and are well suited for on-site screening and large-scale regulatory supervision.¹⁵-²⁰ For instance, NIR spectroscopy combined with chemometric models perfectly discriminates adulterated meat in mutton with 100% validation accuracy, while hyperspectral imaging integrated with independent component analysis provides a promising non-destructive tool for industrial high-throughput beef adulteration screening. Nevertheless, several challenges remain. Model robustness is susceptible to variations in sample morphology, temperature, moisture content, and instrumental parameters.¹⁶ The high cost, instrumental complexity, and data-intensive nature of hyperspectral and Raman equipment restrict their widespread deployment.²³,²⁴ Moreover, high spectral similarity among different meat matrices often causes spectral overlap, necessitating optimized characteristic wavelength selection algorithms to enhance model specificity. Future research should focus on developing portable devices and standardized spectral databases to promote practical industrial implementation.²⁵

3.2. Detection of adulteration in edible oils

Edible oils adulteration refers to the illegal practice of substituting high-grade edible oils with low-cost alternatives for excessive profits, such as blending rapeseed, cottonseed, or sunflower oil with olive oil, or adding palm or soybean oil to camellia oil. Such malpractice disrupts market order and imposes potential health hazards on consumers. With the rapid advancement of spectroscopic techniques, extensive studies have been conducted globally on oil adulteration detection, and spectroscopic methods have gradually demonstrated unique advantages in this field. Table 2 summarizes selected representative studies.

Table 2
Comparison of reported studies on the application of spectral techniques for oil adulteration detection

Detection object	Detection technology	Analytical performance	Reference
Species identification of sesame, soybean, peanut, and corn oils	NIR	identification rate 100%	26
Authenticity identification of rose essential oil	FTIR + Raman spectroscopy	identification rate 100%	27
Percentage of adulteration in pumpkin seed oil and walnut oil	Raman spectroscopy	accuracy rate 95%	28
Adulteration of peanut oil with corn and vegetable oils	NIR + Raman spectroscopy	fast detection, effectiveness	29
Multi - adulteration detection of edible oils	NIR	accuracy reaches 95.8%	30
Adulteration of extra virgin olive oil with different types and concentrations of vegetable oils	HSI + FTIR	FTIR: suitable for high-precision detection; fluorescence HSI: suitable for large-scale detection with efficient batch-processing capability	31

In terms of edible oil identification, spectral data combined with chemometric algorithms enable rapid classification and compositional characterization of oil varieties. Liu et al.²⁶ established qualitative identification models for sesame, soybean, peanut, and corn oils using NIR spectroscopy coupled with cluster analysis, achieving 100% recognition and prediction accuracy. Cebi et al.²⁷ evaluated the performance of FTIR spectroscopy, Raman spectroscopy, and gas chromatography-mass spectrometry (GC-MS) combined with chemometrics in identifying the authenticity of rose essential oil. Hierarchical cluster analysis and principal component analysis (PCA) successfully distinguished authentic commercial rose essential oil from counterfeit samples with perfect accuracy.

For adulteration quantification, spectral feature analysis integrated with chemometrics and machine learning algorithms enables the effective identification and determination of adulterant content. Becze et al.²⁸ adopted Raman spectroscopy combined with PLS regression to quantify adulteration ratios in pure pumpkin seed oil and walnut oil, with the final predictive accuracy reaching 95%. Castro et al.²⁹ performed a comparative qualitative and quantitative analysis of peanut oil adulterated with corn oil and vegetable oil using combined NIR and Raman spectroscopy, demonstrating the effectiveness of vibrational spectroscopy for the rapid detection of peanut oil adulteration. To address the challenge of identifying multi-component adulteration in edible oil, Yuan et al.³⁰ developed a multi-adulteration detection model by integrating one-class PLS variable selection with Euclidean distance kernel function, yielding a detection accuracy above 95.8%. Jiao et al.³¹ compared the performance of fluorescence hyperspectral imaging and FTIR spectroscopy in detecting the adulteration of extra virgin olive oil with vegetable oils of different types and concentrations. FTIR spectroscopy was more suitable for high-precision detection, whereas fluorescence hyperspectral imaging exhibited superior potential for large-scale batch screening due to its high-throughput processing capability.

Spectroscopic techniques coupled with chemometrics deliver high efficiency and precision in edible oil classification and adulteration detection, providing robust technical support for food safety supervision. Continuous technological advances have significantly improved the portability and ease of use of spectroscopic instruments, further enhancing their adaptability in practical applications. The popularization of portable spectrometers enables on-site rapid detection and substantially improves regulatory efficiency.³²,³³ Meanwhile, optimized spectral data fusion algorithms enhance detection reliability and reduce artificial interference.³⁴ Despite remarkable progress, several bottlenecks remain unresolved, including spectral interpretation of complex matrix samples, model universality, and data compatibility across instruments of different brands. Future studies should prioritize technical standardization and methodological optimization to facilitate the large-scale and sustainable application of spectroscopic techniques in food safety governance.

3.3. Honey adulteration detection

As a natural and nutritious food, honey authentication has long been a major research focus in food safety. Although conventional detection methods achieve satisfactory accuracy, they suffer from cumbersome procedures and destructive sample consumption. In recent years, spectroscopic techniques have demonstrated great potential for honey authenticity evaluation due to their rapidity, non-destructive nature, and high efficiency, forming a multi-level research system ranging from basic component analysis to complex adulteration identification.

For rapid quantitative analysis of basic nutritional components in honey, NIR spectroscopy has emerged as a dominant research tool due to its high efficiency and convenience. Anjos et al.³⁵ quantified monosaccharides in 63 honey samples using FTIR-attenuated total reflection (ATR) spectroscopy combined with PLS calibration, employing standard references of trehalose, glucose, fructose, sucrose, melibiose, turanose, and maltose. Qiu et al.³⁶ further expanded the application scope of NIR spectroscopy and established PLS regression models for simultaneous determination of moisture, fructose, glucose, and reducing sugars in honey, with the highest correlation coefficient of each quantitative model reaching 99%. These findings demonstrate that NIR spectroscopy has the potential to replace time-consuming conventional analytical methods, providing reliable technical support for the rapid quality control and online monitoring of honey products.

In the field of honey adulteration identification, research trends feature technological diversification and model refinement. The application of NIR spectroscopy has extended from single-adulterant detection to broad-spectrum identification covering multiple honey varieties and complex adulterants. Tan et al.³⁷ achieved highly accurate honey adulteration recognition using NIR spectroscopy combined with PCA and logistic regression, with accuracies exceeding 98% for both the training and test sets, confirming the promising practical prospects of spectroscopic techniques for rapid and non-destructive detection of honey fraud. Kou et al.³⁸ systematically compared multiple chemometric methods and verified that SVM models can effectively identify honey adulterated with syrup at a level as low as 5%, with discrimination accuracy over 90% for samples with an adulteration ratio of only 1%.

Spectroscopic techniques have become an indispensable analytical tool for honey authenticity research, driving detection methodologies toward rapidity, precision, and practicality in terms of component quantification, complex adulteration discrimination, and specific risk screening.³⁹ Multiple studies have highlighted practical challenges in industrial implementation. Huang et al.⁴⁰ systematically reviewed the merits and limitations of existing detection techniques and pointed out that model universality and instrumental standardization remain urgent issues to be addressed. Khare et al.⁴¹ indicated that incomplete reference spectral databases, the lack of standardized detection protocols, and the high costs of advanced analytical methods limit accessibility for small-scale beekeepers. Interdisciplinary collaboration, improved certification technologies, the construction of open-access reference database, and the development of cost-effective, user-friendly detection tools are essential to ensuring accurate honey authentication. With continuous technological improvement, spectroscopic techniques will play an increasingly vital role in guaranteeing honey quality, regulating market order, and protecting consumer health.

3.4. Detection of dairy product adulteration

Growing consumer demand and frequent occurrence of food safety incidents have created an urgent market need for rapid, non-destructive, and high-efficient detection technologies for dairy quality and safety. Although accurate, conventional analytical methods are constrained by complicated pretreatment, time-consuming operations, sample destruction, and a heavy reliance on large-scale laboratory equipment, failing to meet the requirements of on-site screening and real-time online monitoring. In recent years, representative spectroscopic techniques including NIR, MIR, Raman, and hyperspectral imaging have been extensively investigated and applied in dairy detection due to their unique advantages of rapidity, non-destructive nature, environmental compatibility, and simultaneous multi-component analysis. Table 3 summarizes selected representative studies.

Table 3
Comparison of reported studies on the application of spectral techniques for the detection of milk adulteration

Detection object	Detection technology	Analytical performance	Reference
Milk origin adulteration	Portable NIR	classification accuracy: 98.67%; demonstrating potential for on-site application of portable devices	42
Adulteration of yak milk powder	NIR	adulteration identification and traceability accuracy: 100%	43
Milk freshness	NIR	fast and effective	44
Milk protein content	HIS	R2 = 0.9996	45
Adulteration with melamine and sucrose in milk powder	NIR	correctly distinguished all pure milk powder samples from adulterated samples	46
Melamine and urea in milk powder	FTIR	completed rapid quantification within 1 min, with excellent model linearity	47

In terms of geographical origin tracing and authenticity identification, Zhang et al.⁴² developed a portable NIR spectroscopy method combined with fuzzy uncorrelated discriminant transformation (FUDT) to address geographical adulteration in raw milk, achieving a classification accuracy of 98.67% and demonstrating the great potential of portable spectroscopic devices for field applications. Peng et al.⁴³ focused on high-value yak milk powder and adopted NIR spectroscopy coupled with the KNN algorithm. The proposed method not only successfully identified the adulteration of yak milk powder with conventional milk powder but also realized 100% accurate geographical tracing across four producing regions, including Sichuan and Gansu, establishing an integrated qualitative and quantitative rapid detection framework.

For component quantification and freshness evaluation, Peng et al.⁴⁴ constructed milk freshness discrimination models using the threshold method, linear discriminant analysis (LDA), and SVM. The absorbance of visible-NIR spectra gradually increased with prolonged storage time, accompanied by significant enhancement of autocorrelation and cross-correlation peak intensities in synchronous two-dimensional visible-NIR spectra. The study validated that two-dimensional correlation spectroscopy combined with chemometrics serves as an efficient strategy for rapid milk freshness discrimination. Liu et al.⁴⁵ established an ultra-high precision prediction model for milk protein content using hyperspectral imaging integrated with sparrow search algorithm-optimized (SSA)-SVM, achieving a coefficient of determination as high as 0.9996 and demonstrating the superior performance of hyperspectral technology combined with intelligent algorithms in component quantification.

Targeted detection of illegal additives such as melamine and urea in dairy products has focused on improving detection sensitivity and accessibility. Mazivila et al.⁴⁶ correctly distinguished pure milk powder from samples adulterated with melamine and sucrose using first-derivative NIR spectra coupled with data-driven soft independent modeling of class analogy. Bai et al.⁴⁷ established an FTIR based method enabling rapid quantification of melamine and urea in milk powder within 1 min with excellent model linearity. Tan et al.⁴⁸ developed classification and regression models by integrating NIR spectroscopy with extreme learning machine (ELM), verifying that spectroscopic detection can effectively replace traditional chemical methods for melamine determination in milk.

For complex adulteration and non-targeted screening, methodologies capable of identifying multiple or unknown adulterants possess greater practical significance. Ehsani et al.⁴⁹ explored spectral variations of milk samples via PCA, implemented sample classification using random subspace discriminant ensemble, and quantified water adulteration content via boosted regression trees. Zhang et al.⁵⁰ analyzed goat milk samples adulterated with graded proportions of cow milk using Raman spectroscopy and developed a Gk-neural network regression (Gk-NNR) model to estimate adulteration levels, adopting a stratified Kennard-Stone algorithm for balanced sample partitioning and providing a rapid and efficient approach for goat milk adulteration detection. Huang et al.⁵¹ proposed an autoencoder-based one-class classification method to address the issue of lengthy analytical time in spectroscopic detection. The autoencoder was employed to extract low dimensional features from high-dimensional spectral data and reconstruct original spectra, with reconstruction errors utilized to judge sample adulteration status.

In terms of brand authenticity and economic adulteration identification, Wu et al.⁵² designed a novel fuzzy feature extraction method for clustering milk spectral data, confirming that portable NIR spectrometers can achieve accurate and effective brand classification. Yuan et al.⁵³ established discriminant models for pure and adulterated branded milk powder using visible-NIR spectroscopy combined with the KNN algorithm, demostrating the feasibility of visible-NIR spectroscopy coupled with few-wavelength KNN for high-precision discrimination of milk powder adulteration.

Spectroscopic techniques integrated with advanced chemometrics and machine learning algorithms have shown immense application potential in dairy geographical origin tracing, component quantification, freshness assessment, and targeted / non-targeted adulterant detection. Their inherent characteristics of rapidity, non-destructiveness, and online compatibility perfectly match the demand for efficient quality and safety control in modern dairy industries. Despite prominent advantages, critical challenges persist in practical industrial translation. The adaptability and reliability of laboratory-established models require further validation for enterprise online monitoring and regulatory on-site detection. Future efforts should focus on expanding the scale and representativeness of spectral sample databases and optimizing modeling algorithms and software to enhance model stability and generalizability.⁵⁴

3.5. Condiment adulteration detection

Condiments serve as essential components of the food industry and dietary culture, whose quality and safety are directly linked to consumer health and legitimate rights. Nevertheless, complex chemical compositions and diversified counterfeiting methods pose severe challenges to conventional quality control systems. Traditional sensory evaluation and physicochemical analysis methods are limited by strong subjectivity, time-consuming procedures, sample destruction, and incompatibility with online monitoring. In recent years, rapid and non-destructive analytical techniques, represented by NIR and MIR spectroscopy combined with chemometric models, have become a research hotspot and frontier direction in condiment quality control. Table 4 summarizes selected representative studies.

Table 4
Comparison of reported studies on the application of spectroscopy techniques for the detection of adulteration in condiments

Detection object	Detection technology	Analytical performance	Reference
Active components in star anise (trans-anethole, oxalic acid)	NIR	successfully detected, reducing time consumption	55
Content of α-curcumene in vinegar Atractylodes	NIR	rapid quantification	56
Moisture and acidity during the fermentation process of broad bean paste	NIR	effective identification of moisture and acidity	57
Piperine content in black pepper and performance comparison of portable equipment	NIR	rapid determination of piperine content	58
Umami intensity and overall sensory grade of miso	NIR	crossing from traditional physicochemical indicators to sensory qualities perceived by consumers (such as umami)	59
Authenticity identification of different brands of soy sauce	NIR	high-precision identification of different brands of soy sauce	60
Age identification of aged vinegar with different years (aged vinegar age)	two-dimensional correlation spectroscopy technology	high-accuracy identification of aged vinegar with 1-10 years of aging	61
Consistency, lipid content, and process stability of Qingjiang production line	on-site NIR	real-time monitoring of production process stability and abnormal states	62

The application of spectroscopic techniques has evolved from single-component quantification to comprehensive authentication system covering multiple quality attributes of condiments. Numerous studies have demonstrated that NIR spectroscopy coupled with PLS and other calibration models enables efficient determination of key quality indicators in condiments. For example, Li et al.⁵⁵ and Cao et al.⁵⁶ achieved rapid quantification of active ingredients including trans-anethole and shikimic acid in star anise, as well as α-cyperone in Cyperus rotundus, replacing time-consuming chromatographic methods. Xu et al.⁵⁷ and Mayr et al.⁵⁸ further confirmed the feasibility of spectroscopic techniques in monitoring moisture and acidity during broad bean paste fermentation, quantifying piperine content in black pepper, and even comparing the performance of different portable spectrometers, providing technical solutions for on-site raw material identification. More innovatively, Wang et al.⁵⁹ attempted to directly characterize complex sensory attributes such as umami intensity and overall sensory grade of miso using spectroscopy, promoting the expansion of detection targets from physicochemical indices to consumer-perceived quality characteristics. To address the prevalent malpractice of shoddy substitution and false labeling of geographical origin and production year, researchers have constructed unique spectral fingerprint models by extracting holistic spectral features from samples. The study conducted by Chen et al.⁶⁰ is highly representative; optimized algorithms were employed to screen a small set of critical NIR wavelengths, and the established model achieved high-precision classification of soy sauce from different brands. The few-wavelength modeling strategy provides a theoretical basis for the development of low-cost, dedicated portable identification instruments. Faced with the challenge of high spectral similarity among aged vinegar samples of different years, Zhang et al.⁶¹ innovatively introduced two-dimensional correlation spectroscopy. Taking aging time as an external perturbation, the method significantly improved spectral resolution and distinguished subtle spectral differences among samples, realizing high accuracy identification of vinegar aging ranging from 1 to 10 years and offering a novel technical pathway for year tracing of high-value fermented condiments. With the increasing complexity of adulteration strategies, spectroscopic detection has evolved from targeted analysis of specific additives to global screening strategies based on production process monitoring and non-targeted spectral profiling. Tanzilli et al.⁶² installed online NIR probes on a pesto production line and established multivariate statistical process control models based on real-time spectral acquisition. The system not only determined the consistency and lipid content of final products but also monitored production stability and abnormal operational states in real time. This advancement marks a paradigm shift of spectroscopic identification from offline passive post-inspection to online proactive in-process monitoring, enabling stable and consistent quality control of core fermentation and processing procedures.

Despite fruitful achievements in spectroscopic identification of condiments, large-scale and robust industrial implementation still faces core challenges, particularly regarding model transferability and generalizability. Most current models are established based on limited sample datasets and may suffer dramatic performance degradation when encountering raw materials from different geographical origins, production batches, or slightly adjusted processing technologies.⁶³,⁶⁴ This indicates that further breakthroughs are still required to improve the intelligence, accessibility, and reliability of spectroscopic detection systems for practical industrial deployment.

3.6. Adulteration detection in other high-value food commodities

Spectroscopic techniques, particularly NIR spectroscopy and hyperspectral imaging, are recognized as revolutionary analytical tools for food and agricultural product quality safety detection due to their rapidity, non-destructive nature, and high-throughput performance. Their application scope has extended far beyond conventional dairy and condiment analysis, rapidly penetrating into high-value and high-fraud-risk fields including coffee, Chinese medicinal materials, alcoholic beverages, and functional foods.

As a globally traded commodity, the authenticity and adulteration detection of coffee have attracted widespread attention from academia and industry. Araújo et al.⁶⁵ innovatively integrated NIR spectroscopy with digital image information and adopted one-class classification to achieve non-destructive authentication of premium coffee varieties, achieving 100% recognition accuracy for all samples. Candeias et al.⁶⁶ performed geographical origin certification of instant coffee from southern Bahia using vibrational spectroscopy combined with data-driven soft independent modeling of class analogy. MIR spectroscopy provided detailed chemical composition analysis, while portable NIR spectrometers enable cost-effective in situ detection, offering a robust quality control solution for the coffee industry. Moll et al.⁶⁷ verified that rapid multispectral methods can effectively distinguish closely related Arabica coffee subgroups, with classification accuracies of 94-98% in the NIR region, 88-93% in the visible-NIR region, and 82-93% in the UV-Vis region, demonstrating excellent discriminative potential with room for further methodological improvement.

Chinese medicinal materials and functional foods derived from them, such as Ganoderma lucidum extract, Panax notoginseng, and Poria cocos powder, possess complex compositions and high value, making them prime targets for adulteration and counterfeiting. The application of spectroscopic technology in this field vividly demonstrates its robust capability in addressing complex matrices. Ran et al.⁶⁸ adopted hyperspectral imaging combined with machine learning models for non-destructive detection of chemical components in broken-wall Ganoderma lucidum spores. Hyperspectral data from samples in different producing regions were collected, and back propagation neural network, extreme learning machine, and decision tree models were applied to predict component contents, providing novel insights for the quality evaluation of Ganoderma lucidum products. Yu et al.⁶⁹ found that raw spectra combined with PCA achieved perfect discrimination of Panax notoginseng powder samples, while the identification accuracy for block shaped samples dropped sharply to 9.38% due to physical scattering interference. Systematic comparison of dozens of spectral preprocessing methods confirmed that optimized algorithms such as continuous wavelet transform could elevate the discrimination accuracy to over 93.75%. The critical role of spectral preprocessing was also emphasized in the study of Poria cocos powder by Dong et al.,⁷⁰ who further validated the feasibility of portable spectrometers for rapid on-site detection and promoted the practical popularization of spectroscopic technology.

For special commodities such as alcoholic beverages and chocolate, adulteration poses direct health threats, and spectroscopic techniques exhibit unique advantages in safety screening. Yang et al.⁷¹ acquired visible-NIR spectra of wine using a miniature spectrometer and determined four characteristic wavelengths closely associated with wine adulteration identification through the successive projection algorithm and spectral dimensionality reduction. A four channel spectral sensor was fabricated by integrating a charge-coupled device (CCD) camera with four optical filters, and a random forest classification model was established for adulteration identification. The snapshot-based multispectral sensor achieved a high accuracy of 97.47%, featuring compact size and low cost. Santos et al.⁷² employed NIR and MIR spectroscopy combined with multivariate analytical methods, including PCA and PLS, regression to determine cocoa solid content and detect potential adulteration in chocolate, achieving strong predictive performance with all correlation coefficients exceeding 90%.

Combined with increasingly advanced chemometric and machine learning algorithms, spectroscopic techniques have formed a diversified and high-precision detection system covering geographical origin identification, quality grading, qualitative adulteration discrimination, hazard screening, and precise quantification of high-value commodities including coffee, Chinese medicinal materials, alcoholic beverages, and functional foods. Nevertheless, existing spectroscopic methods still suffer from insufficient detection sensitivity and complete masking of characteristic signals when confronted with complex scenarios involving multiple unknown adulterants. Future research should integrate multi-source data, including spectra, imaging, and electronic nose signals to construct an improved comprehensive identification framework.⁷³,⁷⁴ Spectroscopic techniques will undoubtedly serve as an indispensable technical pillar for building a safer and more transparent market ecosystem for high-value food commodities.

4. Conclusions and Future Perspectives

Spectroscopic analytical techniques, featuring rapid detection, non-destructive measurement, environmental friendliness, online monitoring compatibility and high-throughput screening capability, have evolved into indispensable and transformative tools for food adulteration detection. This review systematically summarizes the state-of-the-art applications of near-infrared spectroscopy, hyperspectral imaging, Raman spectroscopy and mid infrared spectroscopy in adulteration identification of meat, edible oils, honey, dairy products, condiments and other high-value food commodities. Technological advances have driven spectroscopic methodologies from early single-component quantification toward sophisticated system for comprehensive multi-attribute authentication. These techniques enable efficient species identification of meat, variety discrimination of edible oils, and authenticity verification of honey; furthermore, they support geographical origin tracing of dairy products, vintage determination of fermented condiments, and even quantitative evaluation of complex sensory attributes such as umami intensity. Integrated with advanced chemometric and machine learning algorithms, including PLS, SVM, CNN and attention mechanisms, numerous high-precision and robust qualitative classification and quantitative prediction models have been established. Several models achieve accuracies exceeding 95% in independent validation. Strategies involving characteristic wavelength selection, multi-data fusion and imaging visualization have further substantially improved the specificity, intuitiveness and efficiency of spectroscopic detection. These advances firmly validate that spectroscopic approaches serve as efficient and reliable analytical tools, possessing remarkable practical application potential and broad prospects for safeguarding food authenticity, standardizing market order, and protecting consumer interests.

Despite remarkable progress achieved in food adulteration detection, several critical challenges still hinder the large-scale industrial translation of spectroscopic techniques from laboratory research to real-world deployment. First, insufficient model generalizability and robustness remain the core bottleneck. Most existing models are constructed on limited sample datasets confined to specific geographical origins, varieties and processing conditions, thereby suffering dramatic performance degradation when applied to samples from different batches, producing regions or confronted with altered adulteration patterns. Second, interference from complex food matrices and difficulties in low-concentration adulterant detection persist as prominent obstacles. Intrinsic complexity of food compositions, coupled with interference from physical morphology, moisture and fat contents, induces severe spectral overlap and obscures target characteristic signals, leaving detection sensitivity for trace-level and multi-component unknown adulterants in urgent need of further improvement. Third, substantial barriers persist in technological transformation and industrialization. High performance spectroscopic instruments are costly and bulky, while commercially available portable devices still lack satisfactory stability and analytical precision. Moreover, the absence of unified standard spectral databases and consistent model calibration as well as transfer protocols impedes model sharing and reproducibility across studies, further restricting popularization in practical scenarios. Fourth, the heavy reliance on sophisticated algorithms and professional expertise raises technical thresholds, limiting grassroots regulatory adoption and widespread implementation in food manufacturing enterprises.

To facilitate broader and more reliable practical deployment of spectroscopic detection technologies, future research should prioritize the following directions. First, continuous efforts should be devoted to enhancing model robustness and generalization capacity. Large scale standard sample libraries and spectral databases with wide geographical and varietal representativeness need to be established. Meanwhile, artificial intelligence strategies such as transfer learning and meta-learning should be exploited to develop universal or rapidly fine-tunable model architectures adaptable to diverse sample conditions. Second, multi-modal information fusion and intelligent spectral interpretation techniques should be further developed. Integrating spectral data with computer vision, electronic nose / tongue, and even genomic information enables the construction of multi-source collaborative identification frameworks, which overcome the inherent limitations of single spectroscopy and realize comprehensive discrimination against intricate and concealed adulteration behaviors. Third, miniaturization, intelligence and networking of detection equipment should be promoted. Low-cost, high-stability dedicated portable and handheld spectroscopic terminals embedded with edge artificial intelligence (AI) chips ought to be developed to accomplish on-site rapid screening at the terminal side, while cloud platforms undertake iterative model optimization and big data analytics. Fourth, standardization and regulatory system construction require strengthened collaboration among academia, industry and regulatory authorities. Unified full-procedure technical specifications and operational guidelines covering sample preparation, spectral acquisition, model validation and result interpretation should be formulated, and mutually recognized spectral model sharing platforms should be established to pave the way for large-scale commercialization. Addressing these key challenges will undoubtedly elevate spectroscopic techniques from a powerful academic research tool to a normalized, high-efficiency technical pillar underpinning the modern governance system of food safety.

Acknowledgments

The authors gratefully acknowledge the financial support provided by Research Fund Project of Yunnan Provincial Department of Education (2026J0799, 2026Y1085), Yunnan Police College Research Project (202523001034ZR-003, 19A019).

Data Availability Statement

Data sharing is not applicable to this article as no new data were created or analyzed in this study.

References

• ¹ Shi, S.; Zhang, K.; Tian, N.; Food Res. Int. 2025, 6, 211. [Crossref]
  » Crossref
• ² Golinelli, L. P.; Carvalho, A. C.; Casaes, R. S.; J. Dairy Sci. 2014, 97, 6693. [Crossref]
  » Crossref
• ³ Bandiera, A.; Camerlingo, A.; Sanna, N.; Food Control 2026, 179, 111608. [Crossref]
  » Crossref
• ⁴ El-Tahlawy, A. S.; Samkumpim, T.; Wardani, N. I.; Kraiya, C.; Alahmad, W.; Food Cosmet. Toxicol. 2026, 207, 115843. [Crossref]
  » Crossref
• ⁵ Li, F.; Zhang, J.; Wang, Y.; Crit. Rev. Anal. Chem. 2024, 54, 333. [Crossref]
  » Crossref
• ⁶ Mohammadigol, R.; Shahhoseini, R.; J. Med. Plants By-Prod. 2025, 14, 87. [Crossref]
  » Crossref
• ⁷ Lv, H.; Bian, X. H.; Gao, Y.; Wang, M. S.; Wu, H. Y.; Yan, Z. X.; Xiang, Y.; J. Food Qual. 2026, 1, 7348743. [Crossref]
  » Crossref
• ⁸ Ma, J.; Sun, D. W.; Pu, H. B.; Cheng, J. H.; Wei, Q.; Annu. Rev. Food Sci. Technol. 2019, 10, 197. [Crossref]
  » Crossref
• ⁹ Zhu, M.; Huang, D.; Hu, X. J.; Tong, W. H.; Han, B. L.; Tian, J. P.; Luo, H. B.; Food Sci. Nutr. 2020, 8, 5206. [Crossref]
  » Crossref
• ¹⁰ Acharya, B. S.; Nair S.; Salam A.; Compr. Rev. Food Sci. Food Saf. 2026, 25, e70403. [Crossref]
  » Crossref
• ¹¹ Guo, Z. M.; Wu, X. C.; Jayan, H.; Jayan, H.; Xue, S.; Hesham, R.; Zou, X. B.; Food Chem. 2024, 434, 137469. [Crossref]
  » Crossref
• ¹² de Souza, J. C.; Couto, C. D.; Mamede, A. M. G. N.; Valderrama, P.; Toci, A. T.; Freitas-Silva, O.; Eur. Food Res. Technol. 2024, 250, 2639. [Crossref]
  » Crossref
• ¹³ Shi, X. F.; Mod. Chem. Res. 2024, 4, 124. [Crossref]
  » Crossref
• ¹⁴ Feng, L.; Wu, B. H.; Zhu, S. S.; He, Y.; Zhang, C.; Front. Nutr. 2021, 8, 680357. [Crossref]
  » Crossref
• ¹⁵ Zuo, X. J.; Li, Y. L.; Chen, X. W.; Chen, L.; Liu, C.; Processes 2024, 12, 2307. [Crossref]
  » Crossref
• ¹⁶ Oliveira, M. M.; Miliao, G. L.; Rahman, M.; Appl. Food Res. 2026, 6, 101726. [Crossref]
  » Crossref
• ¹⁷ Gao, P. P.; Li, W. L.; Shi, X. Y.; Zhang, X. N.; Zhang, X. N.; Shi, J. Y.; Zou, X. B.; J. Sci. Food Agric. 2026, 106, 2955. [Crossref]
  » Crossref
• ¹⁸ Bai, Z. X.; Zhu, R. G.; Wang, S. C.; Trans. Chin. Soc. Agric. Eng. 2021, 37, 276. [Crossref]
  » Crossref
• ¹⁹ Li, B.; Lu, Y. J.; Liu, Y. D.; Trans. Chin. Soc. Agric. Eng. 2024, 40, 251. [Crossref]
  » Crossref
• ²⁰ Zhai, C.; Li, Y.; Peng, Y. K.; Trans. Chin. Soc. Agric. Eng. 2017, 33, 275. [Crossref]
  » Crossref
• ²¹ Nedeljkovic, A.; Maggiolino, A.; Rocchetti, G.; Sun, W. Z.; Heinz, V.; Tomasevic, I. D.; Djordjevic, V.; Tomasevic, I.; Foods 2025, 14, 3084. [Crossref]
  » Crossref
• ²² Robert, C.; Fraser-Miller, S. J.; Jessep, W. T.; Bain, W. E.; Hicks, T. M.; Ward, J. F.; Craigie, C. R.; Loeffen, M.; Gordon, K. C.; Food Chem. 2021, 337, 128441. [Crossref]
  » Crossref
• ²³ Ahmed, M. T.; Monjur, O.; Khaliduzzaman, A.; Kamruzzaman, M.; Artif. Intell. Rev. 2025, 58, 96. [Crossref]
  » Crossref
• ²⁴ Vignati, S.; Tugnolo, A.; Torrente, M.; Pampuri, A.; Guidetti, R.; Beghi, R.; Giovenzana, V.; J. Agric. Eng. 2025, 56, 1664. [Crossref]
  » Crossref
• ²⁵ Zhang, L.; Trans. Chin. Soc. Agric. Mach. 2020, 51, 478. [Crossref]
  » Crossref
• ²⁶ Liu, C.; Wei, L.; Zhao, W.; China Brew. 2014, 33, 149. [Crossref]
  » Crossref
• ²⁷ Cebi, N.; Arici, M.; Sagdic, O.; Food Chem. 2021, 369, 129495. [Crossref]
  » Crossref
• ²⁸ Becze, A.; Simedru, D.; Not. Bot. Horti Agrobot. Cluj-Napoca 2020, 3, 1426. [Crossref]
  » Crossref
• ²⁹ Castro, R. C.; Ribeiro, D.; Santos, J.; Páscoa, R.; Talanta 2021, 230, 122373. [Crossref]
  » Crossref
• ³⁰ Yuan, Z.; Zhang, L.; Wang, D.; Jiang, J.; Harrington, P. D.; Mao, J.; Zhang, Q.; Li, P. W.; LWT -- Food Sci. Technol. 2020, 125, 109247. [Crossref]
  » Crossref
• ³¹ Jiao, Z.; Song, L. F.; Zhang, Y. L.; Dai, J. W.; Liu, Y. W.; Zhang, Q.; Qin, W.; Yan, J.; J. Food Meas. Charact. 2025, 19, 1761. [Crossref]
  » Crossref
• ³² Li, X.; Wang, D.; Ma, F.; Yu, L.; Mao, J.; Zhang, W.; Jiang, J.; Zhang, L. X.; Li, P. W.; Food Chem. 2023, 405, 134884. [Crossref]
  » Crossref
• ³³ Trindade, T.; Magalhaes, L.; Bandeira, S.; Colares, C.; Acta Sci., Technol. 2024, 46, e63058. [Crossref]
  » Crossref
• ³⁴ Deng, Z. W.; Zheng, Y.; Lan, T.; Zhang, L. X.; Yun, Y. H.; Song, W. R.; Food Chem. 2025, 472, 142930. [Crossref]
  » Crossref
• ³⁵ Anjos, O.; Campos, M. G.; Ruiz, P. C.; Antunes, P.; Food Chem. 2015, 169, 218. [Crossref]
  » Crossref
• ³⁶ Qiu, L.; Liu, Y.; Zhang, Y.; Mod. Tradit. Chin. Med. Mater. Med. World Sci. Technol. 2015, 17, 1949. [Crossref]
  » Crossref
• ³⁷ Tan, S. H.; Pui, L. P.; Solihin, M. I.; Keat, K. S.; Lim, W. H.; Ang, C. K.; J. Food Process. Preserv. 2021, 45, e15576. [Crossref]
  » Crossref
• ³⁸ Kou, Z. K.; Chen, G. T.; Li, S. Y.; Food Sci. 2024, 45, 254. [Crossref]
  » Crossref
• ³⁹ Biswas, A. P.; Tasnim, M.; Süfer, Ö.; Das, S. C.; Sarker, S.; Zhang, M.; Islam, N.; J. Food Meas. Charact. 2026, 20, 3929. [Crossref]
  » Crossref
• ⁴⁰ Huang, J. H.; Fu, Q. Y.; Ren, L. X.; Food Res. Dev. 2025, 46, 196. [Crossref]
  » Crossref
• ⁴¹ Khare, T.; Mosa, K. A.; Hamdy, R.; Elnaggar, A.; Malik, S.; Malik, S.; El-Keblawy, A.; Alhmoudi, A.; Alyammahi, K. M.; J. Food Compos. Anal. 2025, 146, 107974. [Crossref]
  » Crossref
• ⁴² Zhang, T. F.; Wu, X. H.; Wu, B.; Dai, C. X.; Fu, H. J.; J. Food Process Eng. 2022, 45, e14040. [Crossref]
  » Crossref
• ⁴³ Peng, H. Y.; Wu, Z. D.; Lin, T.; Sci. Technol. Food Ind. 2024, 45, 9. [Crossref]
  » Crossref
• ⁴⁴ Peng, D.; Xu, R.; Xu, R.; Yue, J. X.; Su, M.; Zheng, S. S.; Li, J.; Molecules 2023, 28, 5728. [Crossref]
  » Crossref
• ⁴⁵ Liu, M. C.; Xue, H. R.; Liu, J. P.; Spectrosc. Spectral Anal. 2022, 42, 1601. [Crossref]
  » Crossref
• ⁴⁶ Mazivila, S. J.; Páscoa, R.; Castro, R.; Ribeiro, D.; Santos, J.; Talanta 2020, 216, 120937. [Crossref]
  » Crossref
• ⁴⁷ Bai, J.; Wang, H.; Food Ferment. Ind. 2020, 46, 267. [Crossref]
  » Crossref
• ⁴⁸ Tan, C.; Cheng, B.; Chen, M. X.; Liu, W.; Chen, H.; Anal. Lett. 2025, 59, 1337. [Crossref]
  » Crossref
• ⁴⁹ Ehsani, S.; Dastgerdy, E. M.; Yazdanpanah, H.; Parastar, H.; J. Chemom. 2023, 37, e3395. [Crossref]
  » Crossref
• ⁵⁰ Zhang, Y.; Shen, B. E.; Wang, H. Y.; Zhao, Y. J.; Microchem. J. 2025, 215, 114319. [Crossref]
  » Crossref
• ⁵¹ Huang, G.; Yuan, L. M.; Shi, W.; Chen, X.; Chen, X.; Food Chem. 2021, 372, 131219. [Crossref]
  » Crossref
• ⁵² Wu, X. H.; Fang, Y. H.; Wu, B.; Liu, M.; Foods 2023, 12, 3929. [Crossref]
  » Crossref
• ⁵³ Yuan, L.; Chen, X. H.; Chen, X. H.; Chen, J. M.; Pan, T.; Spectrochim. Acta, Part A 2023, 301, 122975. [Crossref]
  » Crossref
• ⁵⁴ Silva, L.; Santos, L.; Ferrao, S.; Int. J. Dairy Technol. 2022, 75, 490. [Crossref]
  » Crossref
• ⁵⁵ Li, Q.; Zhou, Y.; Liu, J.; Chem. Ind. For. Prod. 2023, 43, 38. [Crossref]
  » Crossref
• ⁵⁶ Cao, Q.; Jia, C.; Huang, L.; Chin. J. Exp. Tradit. Med. Formulae 2018, 24, 27. [Crossref]
  » Crossref
• ⁵⁷ Xu, M.; Wang, Y.; Wang, X. B.; Ding, W.; Jia, P. F.; Che, Z. M.; Liu, P.; J. Food Meas. Charact. 2022, 16, 2929. [Crossref]
  » Crossref
• ⁵⁸ Mayr, S.; Bec, K.; Grabska, J.; Schneckenreiter, E.; Huck, C.; Talanta 2020, 223, 121809. [Crossref]
  » Crossref
• ⁵⁹ Wang, S.; Liu, X. F.; Tamura, T.; Kyouno, N.; Zhang, H.; Chen, J. Y.; Anal. Lett. 2021, 54, 2304. [Crossref]
  » Crossref
• ⁶⁰ Chen, J. M.; Fu, C. L.; Pan, T.; Spectrochim. Acta, Part A 2022, 277, 121291. [Crossref]
  » Crossref
• ⁶¹ Zhang, Z.; Han, N.; Zhou, T.; J. Chin. Inst. Food Sci. Technol. 2023, 23, 389. [Crossref]
  » Crossref
• ⁶² Tanzilli, D.; D’Alessandro, A.; Tamelli, S.; Durante, C.; Cocchi, M.; Strani, L.; Foods 2023, 12, 1679. [Crossref]
  » Crossref
• ⁶³ Goyal, R.; Singha, P.; Singh, S. K.; Food Chem. 2025, 477, 143502. [Crossref]
  » Crossref
• ⁶⁴ Yu, X. L.; Zhang, J. H.; Liu, T. A.; Food Mach. 2024, 40, 200. [Crossref]
  » Crossref
• ⁶⁵ de Araújo, L.; Nóbrega, R.; Fernandes, D.; de Araújo, M.; Diniz, P.; da Silva, E.; Food Chem. 2021, 364, 130452. [Crossref]
  » Crossref
• ⁶⁶ Candeias, D.; Silva, K. M.; Pereira, H. S.; Bezerra, L.; Food Chem. 2025, 479, 143698. [Crossref]
  » Crossref
• ⁶⁷ Moll, V.; Bec, K. B.; Handle, L.; Huck, C. W.; Food Front. 2025, 6, 763. [Crossref]
  » Crossref
• ⁶⁸ Ran, J.; Xu, H.; Zhang, W.; Chin. J. Mod. Appl. Pharm. 2025, 42, 4140. [Crossref]
  » Crossref
• ⁶⁹ Yu, M.; Li, S.; Dai, X.; J. Instrum. Anal. 2021, 40, 1374. [Crossref]
  » Crossref
• ⁷⁰ Dong, Y.; Zhao, Z. L.; Fan, W.; J. Food Saf. Qual. 2022, 13, 7210. [Crossref]
  » Crossref
• ⁷¹ Yang, C. B.; Wu, J. J.; Qian, Z. Q.; Zhao, Z. L.; Anal. Methods 2025, 17, 5774. [Crossref]
  » Crossref
• ⁷² Santos, I.; Conceição, D.; Viana, M. B.; Silva, G. D.; Food Chem. 2021, 349, 129095. [Crossref]
  » Crossref
• ⁷³ Mousa, M.; Wang, Y.; Antora, S. A.; Crit. Rev. Food Sci. Nutr. 2022, 62, 8009. [Crossref]
  » Crossref
• ⁷⁴ Jiang, F.; Wu, W. Q.; Li, X.; J. Food Saf. Qual. 2023, 14, 9. [Crossref]
  » Crossref

Edited by

• Editor handled this article:
  Giovanni Wilson Amarante (Executive)

Publication Dates

• Publication in this collection
  17 Aug 2026
• Date of issue
  2026

History

• Received
  18 Mar 2026
• Published
  27 July 2026