Advances in Hyperspectral and Diffraction Imaging for Agricultural Applications

Abstract

Hyperspectral imaging and diffraction imaging technologies, owing to their non-destructive nature, high efficiency, and superior resolution, have found widespread application in agricultural diagnostics. This review synthesizes recent advancements in the deployment of these two technologies across various agricultural domains, including the detection of plant diseases and pests, crop growth monitoring, and animal health diagnostics. Hyperspectral imaging utilizes multi-band spectral and image data to accurately identify diseases and nutritional status, while combining deep learning and other technologies to improve detection accuracy. Diffraction imaging, by exploiting the diffraction properties of light waves, facilitates the detection of pathogenic spores and the assessment of cellular vitality, making it particularly well-suited for microscopic structural analysis. The paper also critically examines prevailing challenges such as the complexity of data processing, environmental adaptability, and the cost of instrumentation. Finally, it envisions future directions wherein the integration of hyperspectral and diffraction imaging, through multisource data fusion and the optimization of intelligent algorithms, holds promise for constructing highly precise and efficient agricultural diagnostic systems, thereby advancing the development of smart agriculture.

1. Introduction

Agriculture serves as the cornerstone of global food security and environmental sustainability. As fundamental pillars of the agricultural system, crop production and livestock management play dual vital roles: ensuring reliable food provision while stimulating rural economic development [1]. Contemporary agriculture confronts escalating challenges from resource limitations, climate variability, and increasing demands for production efficiency. These compounding pressures necessitate transformative innovations in agricultural diagnostics. Modern diagnostic technologies empower precise tracking of crop physiology, livestock welfare, and soil ecosystems, enabling optimized yields and superior resource management. Hyperspectral and diffraction imaging exemplify this technological progression, offering immediate plant disease detection, real-time livestock health monitoring, and early risk identification to safeguard productivity.

These advanced tools generate evidence-based insights for agricultural stakeholders, supporting data-driven policymaking and farm-level decisions that advance sustainable practices. The precision agriculture revolution has consequently amplified the strategic value of diagnostic technologies as essential components of modern agri-food systems [2]. As critical enablers in the agricultural value chain, these imaging technologies enhance both production quality and sustainable practices through improved resource management. Their non-destructive nature, high precision, and operational efficiency position hyperspectral and diffraction imaging as transformative solutions with extensive diagnostic applications across agricultural systems [1,3,4]. Within this context, this review provides (1) a systematic examination of hyperspectral and diffraction imaging fundamentals, (2) an in-depth exploration of their cutting-edge agricultural diagnostic applications, and (3) a critical analysis of current limitations and promising future research directions.

1.1. Hyperspectral Imaging Technology

Hyperspectral imaging represents an advanced diagnostic approach that combines imaging capabilities with spectroscopic analysis. This technique acquires detailed spectral signatures across numerous narrow bands, while simultaneously preserving spatial information [5]. The technology’s principal advantage stems from its dual-output capacity to provide both the spatial distribution and spectral characteristics of targets, enabling precise identification and classification [6]. Hyperspectral imaging captures continuous spectral data across ultraviolet to near-infrared wavelengths (approximately 200–1000 nm), enabling accurate material characterization through detailed optical property analysis [7]. This capability establishes a robust foundation for numerous diagnostic applications [8,9], including classification tasks [10,11,12,13,14], moisture [15,16] and elemental quantification [17,18,19,20,21,22,23,24], colony counting [25,26,27], and sensory assessment [28]. The fundamental principles of hyperspectral imaging are illustrated in Figure 1. A typical hyperspectral imaging system consists of a light source, a sample stage, a spectrometer, and a computer for data acquisition and processing. The light source uniformly illuminates the sample, and the reflected light is captured by the spectrometer. For each spatial point on the sample surface, the system records a complete spectral signature across a wide range of wavelengths (e.g., 400–1000 nm). This results in a three-dimensional hyperspectral data cube containing both spatial (image) and spectral (wavelength) information. Conventional diagnostic methods in agriculture, such as visual inspection, microscopy, and chemical analysis, are often labor-intensive, time-consuming, and destructive. Compared to conventional diagnostic methods, hyperspectral imaging provides distinct advantages, including non-destructive operation [11,29,30,31], rapid data acquisition, high spatial resolution, and comprehensive spectral information content [32,33,34]. These capabilities enable both precise target identification and comprehensive data acquisition for analytical modeling [35]. The advancement of hyperspectral imaging technology stems from dual technological progress: (1) improved sensor hardware performance and (2) sophisticated developments in computational processing methods. The following sections present a comprehensive examination of hyperspectral imaging applications with variant computational algorithms and their diagnostic advantages in agriculture, as well as applications in agricultural science [19,36,37,38,39,40]. As demonstrated in Table 1, these techniques employ diverse analytical approaches, including principal component analysis (PCA), convolutional neural networks (CNNs), support vector machines (SVMs), random forests, graph neural networks, and spectral unmixing methods. Figure 2 presents various types of hyperspectral imaging equipment. Specifically, Figure 2A shows a UAV-based hyperspectral imaging system commonly used for aerial crop monitoring. Figure 2B displays a portable handheld hyperspectral device suitable for the field detection of plant stress. Figure 2C,D illustrate two indoor hyperspectral imaging setups: one for sample scanning under controlled conditions (2C), and another using a fixed frame structure for multi-angle imaging (2D). Figure 2E,F depict two other indoor systems with integrated control modules and optical units for laboratory use. Figure 2G presents a compact handheld hyperspectral camera used for rapid object scanning. Finally, Figure 2H illustrates another UAV-based platform designed for outdoor deployment in agricultural environments.

Figure 1. Basic principles of hyperspectral imaging technology.

Figure 1. Basic principles of hyperspectral imaging technology.

Table 1. Applications of hyperspectral imaging associated with various computational algorithms.

Application Scenarios	Algorithm Type	Advantages	Limitations
Remote sensing classification	Super PCA [41]	Superpixel-based PCA; Preserves spatial structure; Enhances computational efficiency; Unsupervised learning	Computational complexity
	PCA [42]	Effective dimensionality reduction	Potential loss of some important information
	CNN [43]	Generates spectral images from RGB images; Low cost	Resulting image quality depends on training data
	Linear mixing model [44]	Suitable for spectral unmixing; Good model generalization	Model may oversimplify complex spectral variations
	Gabor filter + Unsupervised discriminant analysis [45]	High classification accuracy	High model complexity
	CNN + Dual swin transformer [46]	Captures both local and global features; High classification accuracy	Complex architecture; High training cost; Requires large datasets
Crop image classification	Feature selection + Folded-PCA [47]	Combines information-theoretic optimized feature selection	Implementation complexity
	Information theory-based feature selection + Folded-PCA [48]	Combines information-theoretic feature selection; Enhances classification accuracy	High computational complexity
	CNN + SVM [49]	Combines CNN feature extraction with SVM classification; Adaptable to various crop types	High computational resource demands
	Spatial-spectral homogeneous block extraction [50]	Integrates spatial and spectral information; High accuracy	High model complexity
	CNN [51]	Strong adaptability; High classification accuracy	Long training time
	3D CNN (LeNet-5) [52]	Integrates spatial and spectral information; High classification accuracy	High computational complexity
	GNN + ARMA filter + Parallel CNN [53]	Superior classification performance	High model complexity
	CNN + GAT + C-means [54]	Combines spatial and spectral data; High segmentation accuracy	High model complexity
Soil	Optimal band selection + Random Forest [55]	Improves soil salinity estimation accuracy	Relies on band selection methods
	Evaluation of ML models (SVM, RF, CNN, etc.) [56]	Provides a comprehensive comparison of classification performance	Model accuracy depends on sample quality; affected by spectral interference
	HSI + SVM/RF/PLS-DA [57]	Enables accurate identification of PE and PA microplastics in soil	Feature selection relies on manual design; Limited generalization ability
	HSI combined with radar + PLSR [58]	Reduces soil moisture and surface roughness interference in SOC estimation	Complex fusion process; Requires consistent data sources
	Spectral feature extraction methods (PCA, CARS, GA) [59]	Enhances SOC prediction accuracy through multi-feature integration	High computational cost; Uncertainty in feature selection
	HSI + PLSR + RBF neural network [60]	Enables non-destructive detection of silicon and moisture across regions	Requires retraining for different soil types; Limited model transferability

Table 1. Applications of hyperspectral imaging associated with various computational algorithms.

Application Scenarios	Algorithm Type	Advantages	Limitations
Remote sensing classification	Super PCA [41]	Superpixel-based PCA; Preserves spatial structure; Enhances computational efficiency; Unsupervised learning	Computational complexity
	PCA [42]	Effective dimensionality reduction	Potential loss of some important information
	CNN [43]	Generates spectral images from RGB images; Low cost	Resulting image quality depends on training data
	Linear mixing model [44]	Suitable for spectral unmixing; Good model generalization	Model may oversimplify complex spectral variations
	Gabor filter + Unsupervised discriminant analysis [45]	High classification accuracy	High model complexity
	CNN + Dual swin transformer [46]	Captures both local and global features; High classification accuracy	Complex architecture; High training cost; Requires large datasets
Crop image classification	Feature selection + Folded-PCA [47]	Combines information-theoretic optimized feature selection	Implementation complexity
	Information theory-based feature selection + Folded-PCA [48]	Combines information-theoretic feature selection; Enhances classification accuracy	High computational complexity
	CNN + SVM [49]	Combines CNN feature extraction with SVM classification; Adaptable to various crop types	High computational resource demands
	Spatial-spectral homogeneous block extraction [50]	Integrates spatial and spectral information; High accuracy	High model complexity
	CNN [51]	Strong adaptability; High classification accuracy	Long training time
	3D CNN (LeNet-5) [52]	Integrates spatial and spectral information; High classification accuracy	High computational complexity
	GNN + ARMA filter + Parallel CNN [53]	Superior classification performance	High model complexity
	CNN + GAT + C-means [54]	Combines spatial and spectral data; High segmentation accuracy	High model complexity
Soil	Optimal band selection + Random Forest [55]	Improves soil salinity estimation accuracy	Relies on band selection methods
	Evaluation of ML models (SVM, RF, CNN, etc.) [56]	Provides a comprehensive comparison of classification performance	Model accuracy depends on sample quality; affected by spectral interference
	HSI + SVM/RF/PLS-DA [57]	Enables accurate identification of PE and PA microplastics in soil	Feature selection relies on manual design; Limited generalization ability
	HSI combined with radar + PLSR [58]	Reduces soil moisture and surface roughness interference in SOC estimation	Complex fusion process; Requires consistent data sources
	Spectral feature extraction methods (PCA, CARS, GA) [59]	Enhances SOC prediction accuracy through multi-feature integration	High computational cost; Uncertainty in feature selection
	HSI + PLSR + RBF neural network [60]	Enables non-destructive detection of silicon and moisture across regions	Requires retraining for different soil types; Limited model transferability

Figure 2. Hyperspectral imaging equipment: (A) UAV-based hyperspectral imaging system [61]; (B) Portable outdoor handheld hyperspectral device [62]; (C) Indoor hyperspectral imaging system [63]; (D) Indoor hyperspectral imaging system [64]; (E) Indoor hyperspectral imaging system [65]; (F) Handheld hyperspectral device [66]; (G) Handheld hyperspectral device [67]; (H) UAV-based hyperspectral imaging system [68].

Figure 2. Hyperspectral imaging equipment: (A) UAV-based hyperspectral imaging system [61]; (B) Portable outdoor handheld hyperspectral device [62]; (C) Indoor hyperspectral imaging system [63]; (D) Indoor hyperspectral imaging system [64]; (E) Indoor hyperspectral imaging system [65]; (F) Handheld hyperspectral device [66]; (G) Handheld hyperspectral device [67]; (H) UAV-based hyperspectral imaging system [68].

While offering superior classification accuracy, efficient feature extraction, and robust computational performance, these algorithms confront persistent challenges, including computational intensity, implementation complexity, and stringent data quality requirements.

1.2. Diffraction Imaging Technology

Diffraction imaging is an analytical technique based on light-wave diffraction phenomena. Through the interpretation of diffraction patterns generated by light-object interactions, this method provides high-resolution structural characterization [69]. Its exceptional capability lies in resolving microscopic features with superior sensitivity [70], offering unique advantages for non-destructive investigation of internal material architectures. Diffraction imaging is experiencing rapid growth in modern agricultural diagnostics, attracting significant research interest for its non-destructive evaluation and microscopic analysis capabilities. This technology not only provides a scientific foundation for assessing crop internal quality but also enables livestock health monitoring and detection of subtle compositional variations in animal feed. Continued improvements in hardware performance and computational processing are facilitating the synergistic combination of diffraction imaging with complementary diagnostic approaches, thereby expanding its agricultural applications. The following sections will provide a detailed examination of diffraction imaging’s diagnostic value in agriculture. The principle of diffraction imaging technology is illustrated in Figure 3. Diffraction imaging is a technique that captures and analyzes diffraction patterns resulting from the interaction between light waves, micro-apertures, and the sample. When coherent light illuminates the sample, the sample modulates the incoming wavefront and generates a characteristic intensity distribution pattern. This diffraction pattern is recorded by an image sensor (e.g., color sensor or CMOS) and transmitted to a computer for image reconstruction and feature extraction.

Recent technological progress in high-performance sensors, advanced data processing, and novel algorithms has substantially expanded the agricultural applications of hyperspectral and diffraction imaging [71]. These cutting-edge technologies provide strong support for intelligent agricultural diagnostics and establish a solid technical foundation for sustainable farming practices. However, their widespread implementation still faces significant challenges, particularly regarding high equipment costs and data processing complexity [72]. Future developments through technological innovation and cross-disciplinary cooperation are expected to gradually address these limitations. As these technologies mature, hyperspectral and diffraction imaging are becoming increasingly vital for agricultural applications. Their integration offers significant potential to enhance crop productivity, ensure food security, and support sustainable resource management. This review systematically evaluates recent advances in hyperspectral and diffraction imaging for agricultural diagnostics, with three primary objectives: Comprehensive analysis of technological breakthroughs, critical assessment of existing limitations, and projection of future development pathways. By providing both state-of-the-art synthesis and forward-looking perspectives, this work aims to serve as a foundational reference for researchers while stimulating innovation and practical implementation in the field.

2. Applications of Hyperspectral Imaging Technology in Agriculture

2.1. Disease Detection

Global agriculture continues to experience significant annual crop losses due to plant diseases across all major growing regions [73]. China continues to experience significant annual crop losses due to plant diseases throughout its agricultural regions [74]. This makes timely and precise disease detection essential for protecting crop yields. Current detection methodologies depend predominantly on labor-intensive field inspections and laboratory chemical tests. Although these conventional techniques provide dependable diagnoses, their operational constraints—particularly the extensive time requirements and limited scalability—render them ineffective for regional-scale monitoring [75]. Hyperspectral imaging provides a non-destructive, high-throughput solution for crop disease detection, leveraging its unique combination of high spatial and spectral resolution [76]. The technology acquires detailed reflectance profiles across numerous narrow bands, enabling detection of minute pathological changes in plant physiology and structure. This advanced capability permits identification of pre-symptomatic stress responses, allowing for timely, precision disease management interventions [77,78]. Plant diseases modify leaf optical properties through biochemical and structural changes, including altered chlorophyll content, water status, and cellular organization. These pathological alterations generate diagnostic spectral signatures [79], particularly reduced visible-range reflectance (400–700 nm) from chlorophyll degradation and elevated near-infrared reflectance (700–1300 nm) due to tissue structural damage.

Hyperspectral imaging has become a principal tool for early rice disease detection by capturing these characteristic spectral responses. Feng et al. [80] successfully combined hyperspectral imaging with deep transfer learning to improve rice disease identification. The study acquired hyperspectral images of diseased rice leaves, extracted spectral features, and applied transfer learning algorithms, demonstrating enhanced accuracy. This methodology achieved high classification accuracy despite limited labeled training data, significantly advancing disease monitoring capabilities for rice production systems.

Zhang et al. [81] constructed nine spectral band selection models based on three consecutive years of in situ canopy hyperspectral imaging data to reduce spectral redundancy. Additionally, two novel vegetation indices were developed and combined with a CARS-ridge regression algorithm to identify Fusarium head blight (FHB) in wheat. The results showed that the optimized model achieved a detection accuracy of 91.4%, significantly outperforming traditional methods. The hyperspectral image data used in this study provided high spatial and temporal resolution, which contributed to precise localization of disease spots and enhanced the generalization capability of the model across different regions. Rady et al. [82] employed VIS-NIR hyperspectral imaging (400–900 nm) in push-broom scanning mode to acquire reflectance images of GoldRush apples under different storage temperatures. The reflectance features were classified using five machine learning models: LDA, decision tree, KNN, PLS-DA, and FFNN. Among them, the decision tree achieved the highest overall classification accuracy of 82% on the test set, with class-wise accuracies of 81% for healthy apples and 86% for CM-infested ones. The other models also demonstrated promising performance: LDA achieved 72% overall accuracy, KNN 76%. The decision tree model’s classification accuracy suggests that, in practical non-destructive detection, the model can reliably identify most infested fruit, thereby improving sorting efficiency and reducing manual inspection. However, the residual risk of misclassification could result in a small proportion of infested fruit passing as healthy or healthy fruit being discarded. And forward feature selection (FFS) starts with an empty set and iteratively adds the wavelength that gives the greatest accuracy improvement, repeating until performance gain becomes negligible or a set limit is reached, thus identifying the optimal wavelength subset.

To reduce the number of wavelengths and simplify the model, the authors applied sequential forward selection (SFS) for feature selection. Using only stem orientation data and a significantly reduced number of wavelengths, the SFS-selected subset achieved nearly the same classification performance as the full-wavelength model, with 82% overall accuracy (81% for healthy, 86% for infested). This demonstrates the effectiveness of SFS in dimensionality reduction without compromising detection performance, which is particularly useful for developing compact multispectral systems for practical deployment.

The integration of hyperspectral imaging with deep learning has proven particularly effective for simultaneous classification of multiple crop diseases. Additionally, hyperspectral imaging technology demonstrates strong potential for environmental stress monitoring in crops. Controlled experiments with rice have confirmed its ability to detect pre-symptomatic stress responses, significantly improving both detection accuracy and early warning capacity. These capabilities provide scientifically validated tools for precision agriculture management. Key applications in disease diagnostics are systematically compared in

Table 2. Applications of hyperspectral imaging associated with various target diseases.

Application Type	Target Disease	Detection Technology	Main Achievements
Disease detection	Wheat powdery mildew [83]	HSI	Successfully identified the infection degree of wheat powdery mildew using hyperspectral imaging; Sensitive bands at 560 nm, 680 nm, and 758 nm.
	Potato late blight [84]	UAV-based HSI	Achieved early detection of potato late blight by integrating UAV-mounted hyperspectral sensors with deep learning models.
	Tomato spotted wilt virus [85]	HSI	Successfully detected tomato spotted wilt virus using hyperspectral imaging combined with machine learning techniques; Detection accuracy reached 85%.
	Maize leaf blight [86]	HSI	Achieved high-precision detection of maize leaf blight by integrating hyperspectral imaging with biochemical and spectral features; Overall accuracy reached 86.12%.
	Sucrose diffusion in beef [87]	HSI	Enabled quantitative and spatial visualization of sucrose diffusion dynamics
	Leaf diseases in intercropping [88]	HSI	Successfully distinguished multiple leaf diseases using hyperspectral features
	Rice seeds [89]	HSI	Identified optimal wavebands to improve detection performance and reduce data redundancy
Disease early warning	Potato late blight [90]	Proximal HSI	Early recognition of potato late blight achieved by combining deep learning models with proximal hyperspectral imaging; Test set accuracy reached 73.9%.
	pear leaf anthracnose [91]	HSI	Applied hyperspectral imaging to enable early warning and visual diagnosis of Sclerotinia-infected tomato, improving detection accuracy and providing a basis for timely disease management.
	Citrus leaf diseases [92]	HSI	Summarized the application of multiple hyperspectral imaging technologies in citrus leaf disease detection; Emphasized advantages for fast, non-invasive detection and future development directions.
	Crop seed-borne pathogens [93]	HSI	Demonstrated that hyperspectral imaging combined with AI technology can accurately distinguish between infected and healthy seeds, providing a novel approach for seed pathogen detection.
	Poplar anthracnose [94]	HSI	Developed a spectral model for early and accurate disease detection
Crop remote sensing and monitoring	Apple black rot [95]	Multispectral imaging	Deep learning combined with multispectral imaging enabled early detection of apple black rot, improving orchard disease management efficiency.
	Wheat [62]	UAV-based HSI	Using UAV-based hyperspectral data combined with deep neural network (DNN) analysis of wheat spectral characteristics, achieved yield prediction.
	Rice [80]	HSI	Combined hyperspectral imaging with advanced transfer learning methods to enable rapid detection of various rice upper leaf diseases.
	Soil [96]	Multispectral imaging	By integrating precise spectral features from hyperspectral imaging with dynamic temporal information from multispectral data, significantly improved the accuracy of soil organic carbon estimation.
	Seeds [93]	HSI	Demonstrated that hyperspectral imaging combined with AI can accurately distinguish infected and healthy seeds, providing a novel approach for seed pathogen detection.
	Crop [97]	HSI	Improved classification accuracy and robustness by combining HSI with feature fusion strategies.
Other applications	Fish species [98]	HSI	Proposed a method combining hyperspectral imaging with a SVM model for rapid detection of microplastics in fish intestines, providing a new approach for assessing the impact of environmental pollution on aquatic organism health.
	Honeybees [99]	HSI	Utilized hyperspectral imaging combined with multivariate statistical analysis to detect Varroa destructor mites on honeybee bodies.
	Wheat [62]	UAV-based HSI + Deep learning	Combined UAV-based hyperspectral data with DNN to perform wheat pest and yield prediction, analyzing the correlation between spectral features and pest damage.
	Coffee bean [100]	HSI	Applied hyperspectral imaging to detect damage caused by coffee berry borers, enabling high-precision, non-destructive pest identification in coffee beans.
	Beef, lamb, and chicken [101]	HSI + SVM and ANN-BPN	Vis-NIR HSI outperformed SWIR HSI with higher accuracy (96% vs. 88%), better Rp (~0.99 vs. 0.86), and lower RMSEP (4–9% vs. 15–24%).
	Soybean kernel damage [102]	HSI +RGB+DNN	Achieved high accuracy of 98.36% using spectrum-RGB fusion and optimized convolutional architecture.
	Peaches [103]	HSI + Multivariate analysis	Effectively identified multiple bruises at different stages, enabling improved quality inspection.
	Fruits and vegetables [104]	HSI	HSI enables rapid, non-destructive assessment of moisture, color, damage, and spoilage.
	Microorganism [105]	HSI	Combined modalities accurately predicted microbial growth trends under various packaging conditions.
	Red jujube [106]	HIS + CARS-IRIV and SSA-SVM	Achieved high-accuracy varietal identification through effective feature selection and model optimization.
	Oolong tea [107]	HIS + BOSS-LightGBM	Enabled accurate and rapid classification of tea varieties by combining band selection with advanced machine learning.
	Egg [31]	HSI + SVM	Provided a nondestructive and reliable method for predicting egg freshness with improved regression accuracy.
	Soluble solid content in apples [108]	HSI + DNN	Realized precise prediction of soluble solid content, enhancing nondestructive apple quality evaluation.
	Wolfberry [109]	HSI + Hybrid SVM and LS-SVM	Demonstrated effective differentiation of dried wolfberry quality grades through comparative classification approaches.

.

2.2. Disease Early Warning

Plant diseases cause measurable biochemical and structural modifications in leaves, including changes in chlorophyll content, water status, and cellular organization. These pathological alterations generate distinctive spectral reflectance patterns across the 400–2500 nm range. Hyperspectral imaging detects disease-specific spectral signatures through high-resolution, contiguous spectral band acquisition, allowing identification of minute variations indicative of physiological stress responses. Hyperspectral imaging detects early disease indicators through leaf spectral variation analysis. Terentev et al. [110] acquired hyperspectral data and applied interdisciplinary methods to establish correlations between spectral features and plant biochemical changes. Using SVM classification, the system achieved 97–100% detection accuracy from day 4 post-inoculation, confirming its early-diagnosis potential. Moriya et al. [111] utilized UAV-based hyperspectral imaging (25-band sensor, 500–840 nm range) with concurrent multispectral simulation (3-band) for citrus canker detection and spatial mapping. The hyperspectral imaging system outperformed multispectral analysis in disease classification due to its spectral dimensionality, yielding 0.94 health map accuracy. This methodology has proven effective for both woody (citrus) and monocot crops (cereals), demonstrating versatile agricultural disease monitoring capabilities.

Researchers are increasingly integrating machine learning algorithms with hyperspectral data to improve plant disease identification accuracy. This combined approach, when applied to large-scale agricultural analytics, enhances the reliability of early warning systems for crop disease management. Bai et al. [112] explored hyperspectral imaging for rapid, spatially resolved plant disease monitoring. By analyzing plant tissue spectral features, these methods enable early disease detection and precise outbreak localization, providing a non-destructive tool for agricultural disease management. UAV-based hyperspectral systems further facilitate large-scale field monitoring, allowing timely interventions before widespread outbreaks occur, thus enhancing agricultural sustainability. Xu et al. [113] emphasize multisource data fusion for disease early-warning systems. Theyintegrated hyperspectral images with RGB and thermal infrared data to monitor wheat powdery mildew, demonstrating improved early diagnosis rates and reduced false positives compared to single-data approaches. Representative applications are systematically compared in Table 2. Figure 4 provides representative applications of hyperspectral imaging in early disease detection across multiple crops, such as cotton (A), rice (B), wheat (C), and strawberry (D).

2.3. Crop Remote Sensing and Monitoring

Hyperspectral imaging has become indispensable in crop growth monitoring due to its ability to capture detailed spectral signatures reflecting plant physiological status, including nutrient concentrations and stress responses. The integration of UAV-based hyperspectral systems further enhances this capability by enabling high-resolution, real-time field-scale monitoring, which supports data-driven precision agriculture practices, such as variable-rate fertilization and early-stage disease detection [118]. Recently, numerous research teams have successfully integrated hyperspectral data with machine learning algorithms for the real-time monitoring of crop growth dynamics. Moghimi et al. [62] demonstrated a synergistic combination of UAV-based hyperspectral imaging and DNNs for high-throughput wheat phenotyping and yield prediction. Researchers have employed hyperspectral imagery to collect data from experimental wheat plots and applied deep learning models to extract key features, achieving precise wheat growth stage classification and accurate yield estimation. Furthermore, Wang et al. [119] revealed significant differences in the absorption characteristics of maize leaves in the near-infrared and shortwave-infrared spectral bands during different developmental stages. These spectral variations serve as highly valuable biomarkers for monitoring crop growth dynamics and developmental progress.

Hyperspectral imaging also plays a critical role in assessing crop nutrient status. For example, studies have shown that leaf reflectance varies significantly within specific spectral regions in response to varying nitrogen concentrations. This finding has led to the development of hyperspectral-based nitrogen prediction models, allowing their application in precision fertilization strategies to optimize agricultural productivity. Additionally, UAV-based hyperspectral imaging has been used to monitor chlorophyll density in maize canopies subjected to lodging stress. In one study, controlled experiments were conducted to identify sensitive spectral parameters for modeling. The results revealed that increased stem reflectance, compared to leaf reflectance, was the primary driver of spectral variation under lodging conditions. Furthermore, vegetation index-based models exhibited the highest predictive accuracy, enabling effective classification of maize lodging severity [120]. Hyperspectral imaging remote sensing has demonstrated significant potential for crop type discrimination, substantially improving the precision of agricultural census data. Recent research has explored the integration of remote sensing data with deep transfer learning for crop classification. A particularly innovative approach involves adapting pre-trained CNNs—initially designed for general image recognition—to process multispectral satellite imagery for precise crop type identification across diverse agricultural systems. This methodology has proven particularly effective in overcoming the constraints of limited annotated training data, thereby enhancing the scalability and accuracy of crop monitoring programs [121]. Table 2 provides a comprehensive overview of hyperspectral imaging applications in agricultural remote sensing and surveillance.

2.4. Extended Applications of Hyperspectral Imaging Techniques

Hyperspectral imaging has emerged as a powerful tool for animal health assessment and disease diagnostics. This technology enables non-invasive differentiation between healthy and pathological tissues through spectral signature analysis of animal surfaces or organs. A notable application includes the rapid identification of microplastic contamination in fish intestinal tracts, demonstrating its utility in assessing environmental health impacts on aquatic species [98]. Furthermore, hyperspectral imaging facilitates early detection of livestock infections by analyzing spectral variations in skin or hair associated with parasitic or bacterial pathogens, thereby supporting prompt clinical intervention. In addition to the applications summarized in Table 2, hyperspectral and multispectral imaging technologies have also been utilized in the tea industry for quality evaluation and fermentation monitoring, as illustrated in Figure 5. Panel A depicts the spectroscopic acquisition of spectral and image data from tea samples, whereas panel B outlines an integrated workflow combining multispectral imaging, chemical index acquisition, data augmentation, PLS- and CNN-based modeling, and pixel-level visualization to monitor the dynamic fermentation process of black tea.

In comparison with conventional destructive techniques and visual inspection, hyperspectral imaging integrated with machine learning offers a non-invasive, early-stage warning, and accurate approach for detecting pest infestations. This technique is capable of identifying internal physiological changes prior to the manifestation of visible symptoms. Nevertheless, several limitations persist, including the high cost of instrumentation, the complexity of data preprocessing, and susceptibility to variations in surface conditions. In the future, detection accuracy could be improved by enlarging the dataset to include more varieties and orchard sources, refining wavelength selection and spectral preprocessing, and adopting ensemble methods such as random forests or gradient boosting, combined with cross-validation to reduce overfitting.

3. Applications of Diffraction Imaging Technology in Agriculture

Microstructural analysis plays a pivotal role in agricultural research, with critical applications spanning crop growth assessment, cellular characterization, pathogen identification, and postharvest quality control. Conventional approaches, including optical microscopy, scanning electron microscopy (SEM), and laser confocal microscopy, offer high-resolution imaging capabilities but are constrained by demanding sample preparation protocols and substantial operational costs, particularly for field-deployable diagnostics. Recent advances in diffraction-based imaging technologies, especially lens-free configurations, have introduced transformative capabilities for agricultural microanalysis. These systems provide distinct advantages through their lensless design, operational simplicity, and capacity for high-throughput processing, making them particularly suitable for in-field applications (Figure 6). Figure 6A illustrates lens-free imaging for real-time cell viability monitoring, where phase contrast generated by light diffraction is used to distinguish between live and dead cells. Figure 6B depicts a lens-free imaging system designed for capturing transient processes, such as rapid biochemical reactions, by combining microfluidics with a compact optical setup. Figure 6C demonstrates a high-throughput lens-free microscopic imaging platform, which employs a diffractive optical element and time-sequenced illumination to extract dynamic spectral data across frames. Figure 6D shows an application of lens-free imaging in fungal spore detection, where pathogenic fungi generate unique diffraction fingerprints on a CMOS sensor, enabling precise identification and classification of different fungal species through computational reconstruction.

3.1. Classification and Identification of Pathogenic Spores

In the transmission of crop diseases, the production and dispersion of pathogenic spores represent one of the most critical early indicators. As vectors of various fungal diseases, spores are often present in the air or on plant surfaces before visible symptoms appear. Therefore, the detection and identification of spores are considered key steps in early warning and precise control of plant diseases. Especially in protected agriculture and open-field farming environments, the rapid and high-throughput capture and recognition of airborne or surface-bound spores are essential for early intervention. Compared with conventional single-wavelength diffraction imaging techniques, the lens-free diffraction imaging system enhances the differentiation of morphologically similar spores by incorporating multi-wavelength illumination and defocus-based diffraction acquisition. Due to differences in pigment distribution or internal composition, different types of spores exhibit distinct absorption characteristics at specific wavelengths. These variations enable morphologically similar spores to display distinguishable features through the combination of diffraction and spectral information, thereby significantly improving classification accuracy and robustness.

Recent years have witnessed significant technological progress in spore detection [128,129], with numerous innovative methodologies being developed to enhance agricultural disease management. Particularly in rice pathogen surveillance, microfluidic-integrated diffraction reconstruction techniques have emerged as a prominent research focus [130,131,132,133,134,135,136]. Yang et al. [137] developed a novel rice blast spore detection method utilizing diffraction image light field analysis combined with advanced texture feature extraction. Comparative experiments revealed superior identification accuracy compared to conventional manual microscopy. Following neural network optimization, the system demonstrated rapid diagnostic capability, achieving spore classification within seconds while maintaining 97.18% test accuracy. This breakthrough represents a significant advancement in field-deployable phytopathogen detection technology. In a complementary approach, researchers have implemented a microfluidic-based purification system for airborne spores. Through meticulous optimization of chip architecture, material selection, and fabrication protocols, the team successfully overcame key challenges, including environmental contaminant interference and low target spore concentrations [138]. For greenhouse crop applications, the integration of diffraction–polarization imaging signatures with machine learning algorithms has demonstrated particular efficacy. A specialized imaging system was developed to acquire spore diffraction fingerprints across multiple polarization states. Subsequent feature extraction and classification using an SVM algorithm yielded exceptional taxonomic discrimination among prevalent phytopathogens: Botrytis cinerea (tomato gray mold), Pseudoperonospora cubensis (cucumber downy mildew), and Podosphaera xanthii (cucumber powdery mildew). The system achieved a mean classification accuracy of 95.85%, establishing a robust framework for automated spore identification and categorization [139].

To overcome the limitations of conventional diffraction methods in discriminating morphologically similar spores, researchers have pioneered large-field, lens-free, multispectral diffraction recognition systems. Through the development of customized optical configurations coupled with advanced computational algorithms, these innovative techniques achieve precise taxonomic differentiation of various spore species. Customized optical configurations play a critical role in enhancing the accuracy of spore taxonomic identification. By tailoring the optical path—including the illumination wavelength range, angular uniformity, and sample-to-sensor distance—the system maximizes the resolution and contrast of diffraction patterns. These adjustments allow for clearer visualization of species-specific fringe details that may otherwise be lost in standard setups. Furthermore, spectral band selection is optimized to highlight subtle differences in the spores’ optical responses, improving the separability of morphologically similar species. This targeted enhancement of optical contrast directly contributes to improved classification performance and model robustness in downstream machine learning analysis. Meanwhile, advanced computational algorithms play a pivotal role in decoding multispectral diffraction data for precise spore classification. Given that diffraction fingerprint images are often visually indistinguishable due to their complex and subtle patterns, manual analysis is impractical. Each spore species, however, exhibits unique spectral–diffractive responses across different illumination wavelengths. Machine learning models, particularly convolutional neural networks (CNNs), are capable of automatically extracting discriminative features from these high-dimensional datasets. This highlights the pivotal role of advanced computational algorithms in processing multispectral diffraction data for accurate spore classification. By integrating spectral and spatial information, these algorithms enhance class separability and classification robustness. Additionally, techniques such as spectral normalization, dimensionality reduction, and multi-channel fusion further contribute to improved model performance and generalization. The integration of lens-free multispectral diffraction technology improves the scalability of agricultural pathogen surveillance by offering a compact, cost-effective, and easily deployable solution. Its lensless architecture eliminates bulky optical components, simplifying the system’s structure and enabling miniaturization. This makes the platform highly adaptable to field environments such as greenhouses, open farms, and storage facilities. Furthermore, its compatibility with automated sampling devices and microfluidic modules facilitates high-throughput and distributed deployment, supporting large-scale pathogen monitoring in real-world agricultural settings [140]. Spore detection based on diffraction imaging techniques is summarized in

Table 3. Applications of different imaging associated with various target diseases.

Application Type	Target	Detection Technology	Main Achievements
Spore detection	Rice disease spores [141]	Microfluidic chip combined with diffraction-based optical technology	Optimized microfluidic chip structure to achieve purification and detection of low-concentration airborne spores, significantly improving detection sensitivity.
	Fungal spores [139]	Diffraction–polarization imaging combined with machine learning	Developed a method based on diffraction-polarization imaging features of fungal spores; Integrated with SVM algorithm to achieve high-accuracy identification of multiple fungal spores.
	Rice virus spores [133]	Portable microfluidic chip combined with lens-free diffraction imaging	Proposed a portable device using microfluidic chip and lens-free diffraction imaging for rice virus spore capture and detection; Results highly correlated with microscopic identification.
	Rice blast spores [137]	Diffraction-based texture feature analysis	Developed a rapid rice blast spore detection and identification method based on diffraction texture analysis, enabling effective classification of rice blast spores and other spores.
	Airborne crop pathogen spores [142]	Diffraction-based optical identification sensor network	Built a diffraction-based optical identification sensor network for airborne crop pathogen spores, enabling source localization and monitoring, and enhancing disease early warning capabilities.
	Pathogenic fungal spores [127]	Multispectral diffraction-based microfluidic sensing system	Developed a portable multispectral diffraction-based microfluidic sensing system for pathogenic fungal spore detection; Demonstrated high sensitivity and specificity.
Pathogenic spore counting	Spores (tomato gray mold, cucumber downy and powdery mildew) [139]	Diffraction–polarization fingerprint + SVM	~95.85% accurate classification of multiple spore types using diffraction–polarization texture features.
	Greenhouse crop fungal spores [140]	CMOS sensor + Diffraction fingerprint processing	Compact, high-throughput detection with ~92.7% SVM accuracy.
	Rice disease spores in microfluidic chip [136]	Microfluidic capture + Lensless diffraction	Portable system enabling automatic spore detection via diffraction fingerprint imaging.
	Aspergillus niger spore sporulation [143]	Magnetic nanoparticle exposure + Spore count analysis	Demonstrated that larger magnetic nanoparticles significantly inhibit spore production.
Other applications	Microorganisms in sediment [144]	Lens-free imaging technology	Used lens-free imaging to capture images of microorganisms in sediment; Analyzed morphological and physiological characteristics to achieve microorganism species identification.
	Virus-infected cells [145]	Lens-free diffraction imaging technology	Analyzed diffraction pattern signatures of virus-infected cells; Achieved high-throughput, visual monitoring with a linear correlation of 98.9%.
	Virus-infected cells [146]	Lens-free imaging technology	Proposed a novel method using lens-free diffraction pattern analysis for biological detection of virus-infected cells, enabling efficient, flexible identification and differentiation of various viruses.
	Cyst nematode eggs in soil samples [147]	Lensless or deep learning-based counting	Greatly improved egg purity and automated counting efficiency with minimal manual error.
	Rice blast fungal spores [137]	Lensless diffraction imaging + CNN	Enabled rapid (few seconds) identification of fungal spores with 97.18% accuracy.
	Fluorescently labeled particles or cells [148]	Lensless fluorescence imaging + Hybrid bandpass filters	Achieved wide field-of-view imaging with high signal-to-noise ratio using a compact, low-cost system.
	Cells and microorganisms [149] Cells and microorganisms [150]	Lensless imaging + Deep learning analysis	Enabled automated identification of cell morphology with high resolution in a portable format.
		Lensless holographic imaging	They present a compact lensless holographic imaging system featuring auto-focusing and deep learning-based detection for label-free cell classification.

.

3.2. Pathogenic Spore Counting

Spore counting constitutes a fundamental procedure in microbiological and agricultural research. Conventional methodologies predominantly rely on microscopic enumeration and culture-based approaches. The microscopic counting technique typically utilizes a hemocytometer, where a spore suspension aliquot is loaded into the counting chamber. Spores within predetermined grid regions are then visually quantified under optical magnification, with the total concentration derived through dilution factor calculations. Although this approach offers procedural simplicity, it exhibits significant susceptibility to observer-induced variability and demands substantial technical proficiency to maintain counting precision. Alternatively, culture-based quantification involves inoculating spore samples onto selective growth media. Subsequent incubation under controlled environmental parameters enables colony formation, with the enumerated colony-forming units serving as a proxy for initial spore concentration. This methodology provides the distinct advantage of discriminating between metabolically active and inactive spores. Nevertheless, these techniques necessitate extended processing durations and rigorous maintenance of cultivation parameters to ensure reproducible outcomes.

In recent years, diffraction imaging—particularly lens-free diffraction imaging—has emerged as a promising technique for spore quantification. This approach enables rapid and accurate spore counting through analysis of characteristic diffraction patterns. Zhang et al. [151] presented a crop disease monitoring system that utilizes airborne spore sensors based on diffraction light scattering patterns to identify and track pathogen sources. By analyzing the unique diffraction signatures of fungal spores in real time, the network enables precise disease source localization and field-scale surveillance, offering a novel approach for early disease detection and management in agricultural fields. This system acquires high-resolution, real-time images of individual spores and precisely quantifies spore counts. By integrating with an Internet of Things (IoT) platform and incorporating environmental data from multiple sensors, the system facilitates real-time disease evaluation and delivers actionable, high-accuracy monitoring information to farmers. Together, these studies demonstrate the notable benefits of diffraction imaging for spore enumeration and its potential as a reliable tool for early detection and prevention of agricultural diseases.

In a greenhouse environment, a method for the rapid detection of Botrytis cinerea spores in tomatoes was proposed based on microfluidic chip enrichment and lensless diffraction image processing. In this approach, airborne spores were effectively concentrated using a microfluidic device, and their diffraction patterns were captured by a lensless imaging system [136]. Image processing algorithms were then applied to analyze the spore information, enabling rapid identification and detection. Although this study did not specifically conduct quantitative spore counting experiments, its image acquisition and processing framework demonstrates strong scalability, providing both theoretical and technical support for future applications in spore concentration estimation and automatic counting. In addition, several other studies, while not directly aimed at spore quantification, have successfully achieved species-level spore classification using diffraction texture features or polarization fingerprint images. These findings suggest that combining diffraction image texture extraction techniques with deep learning-based counting networks may allow simultaneous identification and quantification of spores from a single image, thus advancing spore detection toward higher precision and automation. Related studies are summarized in the ‘Other applications’ section of Table 3.

3.3. Cell Viability Detection

Cell viability detection holds significant importance in agricultural microbiology, pesticide toxicity assessment, and the development of biocontrol agents. Cell viability directly impacts the efficacy of microbial fertilizers and the precision of pesticide toxicity evaluations. Viability, as a biological indicator, reflects the collective behavior of large cell populations and thus requires statistical analysis of high-throughput sample data for accurate assessment. Traditional detection methods—such as the MTT colorimetric assay, trypan blue staining, and flow cytometry—offer high sensitivity and accuracy. However, these techniques typically require complex sample preparation, expensive instruments, or chemical reagents, which limits their applicability in field-based diagnostics. In recent years, diffraction imaging-based methods for assessing cell viability have attracted growing research interest. In particular, lens-free diffraction imaging technology has demonstrated significant potential for evaluating the viability of agricultural pathogenic microorganisms. Its advantages include label-free operation, low cost, and high-throughput capacity, which makes it well-suited for on-site viability assessment in agricultural settings [152].

Recent studies have leveraged diffraction-based technologies to develop high-throughput, lens-free three-dimensional (3D) imaging techniques capable of label-free, high-throughput 3D tracking of over 1500 sperm cells simultaneously. These investigations revealed that only 4% to 5% of motile sperm exhibited characteristic helical swimming patterns, a behavior found to be inhibited by seminal plasma. Moreover, approximately 90% of helical sperm displayed right-handed helices, and key motility parameters such as helix radius were quantified, providing a powerful high-throughput tool for microbiological research [153]. Furthermore, Li et al. [126] proposed two quantitative parameters—fringe intensity contrast (FIC) and fringe dispersion (FD)—to describe the clarity and spatial distribution of cell diffraction patterns. It was observed that increasing methylmercury concentrations led to a marked decrease in FIC and an increase in FD, reflecting a loss of membrane integrity and fringe pattern degradation, thereby indirectly indicating reduced cell viability. Compared to conventional biochemical assays requiring staining or labeling, this diffraction-based approach enables non-invasive, real-time monitoring of live cells. BRL cells were exposed to different concentrations of methylmercury, and the researchers measured the fringe intensity contrast (FIC) and fringe dispersion (FD) parameters for each cell using a self-developed lens-free diffraction imaging system. In parallel, the classical MTT colorimetric assay was used to assess cell viability. The results from both methods were then compared to evaluate their consistency in assessing cell viability under varying methylmercury exposures. The findings showed that the FIC and FD parameters exhibited high consistency with MTT results, confirming their reliability in reflecting cell viability. However, the generalizability of this technique to other cell types may be limited by factors such as cell size, morphological consistency, and diffraction signal strength, suggesting the need for further standardization and image processing enhancement. Notably, this technique eliminates the need for labeling agents and avoids the requirement for periodic biochemical reaction steps, thereby improving the accuracy and efficiency of cell viability monitoring.

Meanwhile, studies have employed high-throughput lensless imaging technologies to monitor heterogeneous cell suspensions, providing valuable insights for agricultural applications such as cell counting and activity assessment. Su et al. [154] proposed a chip-based platform utilizing lens-free diffraction imaging (LUCAS), which is capable of capturing diffraction patterns of multilayered cells within a 4 mL solution across a field of view of approximately 10 cm² and a depth of 4 mm, all within less than one second. The integrated algorithm enables simultaneous 3D localization and classification of cells. The platform’s automation and high-throughput capabilities offer technological inspiration for the concurrent analysis of cell viability and population in agricultural microbial fertilizers. Drawing from the imaging and analytical strategies of such lensless diffraction systems, there is considerable potential for broader applications in agriculture, including plant tissue culture and seed viability assessment. For instance, in plant tissue culture, this methodology could be adapted for non-destructive viability screening of cell clusters, thereby improving the stability of the cultivation system [154]. In seed screening, the diffraction texture patterns generated by active microcells within the embryo may serve as a promising non-contact indicator of germination potential [155]. These cross-disciplinary applications highlight the bridging potential of diffraction imaging technology between agricultural fundamental research and field-based applications.

3.4. Other Agricultural Applications of Diffraction Imaging Technology

Lens-free diffraction imaging technology offers unique advantages in agricultural microbiological detection. For example, Jiang et al. [144] proposed a microbial species identification method based on colony fingerprinting. Using lens-free imaging, researchers captured images of microcolonies (with diameters ranging from 10 to 500 μm) and distinguished different microbial species by analyzing the morphological and physiological characteristics present in these images. Beyond microbial detection, this technology also shows transformative potential in livestock and poultry health monitoring. Chen et al. [156] introduced a particle separation and detection system that integrates microfluidics with lens-free imaging. The system employs standing surface acoustic waves (SSAW) to achieve efficient particle separation, followed by real-time detection of the separated particles using lens-free imaging. Compared to conventional PCR-based methods, lens-free diffraction imaging offers greater convenience, making it ideally suited for large-scale, field-based monitoring applications.

Lens-free diffraction technology offers broad and far-reaching applications. In addition to its uses in the domains discussed above, it also plays a critical role in fields such as biomedical research and materials science and demonstrates tremendous potential in areas including environmental monitoring and food safety. These capabilities provide strong technological support for advancements across a wide range of industries. Moreover, diffraction technology holds significant promise in animal disease detection. For instance, in the context of veterinary disease prevention, researchers have employed this technology to monitor the infection process of the pseudorabies virus (PRV) in swine. By analyzing the optical diffraction fingerprint patterns of cells during viral infection, they achieved high-throughput, real-time visualization of the infection process. An online, location-based detection system based on lens-free diffraction imaging was also developed. This method exhibited a linear correlation of 98.9% with conventional “gold-standard” diagnostic techniques and offers key advantages such as high automation, high throughput, and low cost. As such, it provides a novel avenue for viral disease prevention and presents a new perspective for cell-level selection and breeding of superior livestock and poultry strains [146].

4. Conclusions

Both hyperspectral and diffraction imaging deliver distinct advantages for agricultural diagnostics, delivering efficient, non-destructive solutions across disease monitoring, crop growth analysis, agricultural product quality assessment, and pathogen detection. Capitalizing on their complementarity, recent studies are pioneering the integration of these technologies. For instance, a novel image reconstruction framework fuses diffraction-induced rotationally blurred images and sharp reference images to achieve high-quality hyperspectral image reconstructions [157]. Jeon et al. [158] introduced a hyperspectral imaging system that integrates diffraction optics with rotational snapshot imaging, aimed at addressing the limitations of conventional hyperspectral systems in terms of imaging speed and device size. This system replaces traditional dispersive components, coded apertures, and relay lenses with diffraction optical elements (DOEs), thereby achieving a more compact design and significantly faster imaging. Additionally, a novel multispectral snapshot diffraction-based metrology technique was presented, wherein the linear relationship between off-diagonal elements of the Mueller matrix and stacking errors was analyzed. This enabled a breakthrough in achieving wide-spectrum measurements in a single exposure—dramatically improving both measurement speed and system stability compared to traditional methods. These innovations demonstrate cross-domain potential, with applicability extending to nanoscale overlay metrology in semiconductor manufacturing [159].

Although these initial integration efforts are promising, several bottlenecks remain in practical applications [160]. The widespread adoption of these technologies continues to be constrained by factors such as the complexity of data processing, equipment costs, and environmental adaptability. Future research must therefore prioritize the following:

4.1. Integration of Hyperspectral and Diffraction Imaging

To date, relatively few studies have explored the deep integration of hyperspectral and diffraction imaging technologies. Future research may adopt multimodal data fusion strategies to combine the rich spectral information of hyperspectral imaging with the microstructural analysis capabilities of diffraction imaging, thereby enhancing diagnostic precision. For example, in disease diagnostics, hyperspectral imaging may facilitate large-scale screening, while diffraction imaging could further identify pathogenic spores, facilitating precise, multiscale detection.

4.2. Advanced Data Analytics and Artificial Intelligence

Both hyperspectral and diffraction imaging generate vast volumes of data, and traditional analytical methods incur high computational costs, limiting their feasibility for real-time agricultural monitoring. Future work could leverage deep learning and machine learning algorithms to optimize data processing and improve both accuracy and speed. For instance, multimodal analysis models combining CNNs with attention mechanisms could be developed to enhance early disease recognition capabilities.

4.3. Development of Low-Cost, Portable Devices

At present, the high cost of hyperspectral and diffraction imaging equipment constrains their application in small- to medium-scale agricultural settings. Future efforts must prioritize the development of portable, low-cost sensing devices that integrate optoelectronic sensors, MEMS technologies, and mobile computing processing platforms to improve accessibility. For example, compact hyperspectral diffraction imaging systems mounted on small UAVs may facilitate real-time field monitoring.

4.4. Enhanced Environmental Adaptability and Robustness

Field environments are highly variable, with factors such as lighting, temperature, and humidity affecting imaging quality. Future research could employ adaptive optical correction and environmental factor modeling to improve system stability. For instance, integrating spectral correction algorithms could mitigate the influence of ambient lighting on hyperspectral data, thereby enhancing detection reliability.

4.5. Interdisciplinary Integration and Smart Agriculture Applications

The advancement of hyperspectral and diffraction imaging technologies requires deep integration with disciplines such as agricultural science, materials science, computer vision, and robotics, to drive the development of intelligent agriculture. For example, intelligent agricultural decision-support systems could be designed to integrate multisource sensor data and enable precision agricultural management.

5. Future Prospects

Hyperspectral and diffraction imaging technologies hold significant promise for advancing agricultural diagnostics. However, key technical challenges remain, including equipment costs, data processing complexity, and environmental adaptability. In addition to these general challenges, deploying lens-free multispectral diffraction systems in practical agricultural settings presents specific obstacles. Variations in lighting conditions, airborne particulates, and humidity can compromise image consistency. Moreover, real-world samples often contain non-target contaminants, making it difficult to isolate spores without upstream filtration. Ensuring long-term operational stability and preventing optical misalignment or sensor fouling during field use are critical concerns that must be addressed for reliable deployment. Looking forward, the integration of multimodal data fusion, artificial intelligence-driven optimization, development of low-cost portable devices, and incorporation into intelligent agricultural systems will further enhance the precision and intelligence of agricultural diagnostics. These advancements will contribute to the sustainable development of modern agriculture.