Search Results (3,604)

Citrus fruits, classified under the Rutaceae family and Citrus genus, are valued for their high nutritional content, attributed to their rich array of natural bioactive compounds. To ensure both quality and nutritional value, precise non-destructive testing methods are crucial. Among these, computer vision and spectroscopy technologies have emerged as key tools. This review examines the principles and applications of computer vision technologies—including traditional computer vision, hyperspectral, and multispectral imaging—as well as various spectroscopy techniques, such as infrared, Raman, fluorescence, terahertz, and nuclear magnetic resonance spectroscopy. Additionally, data fusion methods that integrate these technologies are discussed. The review explores innovative uses of these approaches in Citrus quality inspection and grading, damage detection, adulteration identification, and traceability assessment. Each technology offers distinct characteristics and advantages tailored to the specific testing requirements in Citrus production. Through data fusion, these technologies can be synergistically combined, enhancing the accuracy and depth of Citrus quality assessments. Future advancements in this field will likely focus on optimizing data fusion algorithms, selecting effective preprocessing and feature extraction techniques, and developing portable, on-site detection devices. These innovations will drive the Citrus industry toward increased intelligence and precision in quality control. Full article

Rice seeds’ infection with rice blast will directly lead to rice yield reduction or even crop failure in the next year. Therefore, it is very important accurately identify infected rice seeds. In this study, deep learning and hyperspectral imaging techniques were used for that purpose. First, hyperspectral image data were collected. Then, the UeAMNet (unsupervised extraction attention-based mixed CNN) model—designed in this study—was used to analyze these data and the results compared with the 2DCNN, 3DCNN, A2DCNN, A3DCNN, Ue2DCNN, Ue3DCNN, UeA2DCNN, UeA3DCNN, MNet, AMNet and UeMNet models using different training set (Tr) sizes. The results showed that the new UeAMNet model was superior to the comparison models when using different Tr sizes, and the accuracy could reach 100%. Notably, when Tr was only 0.05, the accuracy of this model still reached 96.85%. This showed that the proposed method could successfully identify infected rice seeds. Therefore, this study provides an approach for rice germplasm management and also for the development of crop disease identification methods in other parts of the world. Full article

The pH value of maize silage can accurately reflect its quality. In this study, a colorimetric sensor array (CSA) combined with hyperspectral imaging (HSI) was used to predict the pH value of maize silage during secondary fermentation. Seventeen color-sensitive dyes were used to construct the CSA, which was subsequently applied to capture the volatile odor profiles of maize silage samples. Hyperspectral images of the color-sensitive dyes on the CSA were acquired using the HSI technique. Different algorithms were used to preprocess the raw spectral data of each dye, and a partial least squares regression (PLSR) model was built for each dye separately. Subsequently, the adaptive bacterial foraging optimization (ABFO) algorithm was employed to identify three color-sensitive dyes that demonstrated heightened sensitivity to pH variations in maize silage. This study further compared the capabilities of individual dyes, as well as their combinations, in predicting the pH value of maize silage. Additionally, a novel feature wavelength extraction method based on the ABFO algorithm was proposed, which was then compared with two traditional feature extraction algorithms. These methods were combined with PLSR and backpropagation neural network (BPNN) algorithms to construct a quantitative prediction model for the pH value of maize silage. The results show that the quantitative prediction model constructed based on three dyes was more accurate than that constructed based on an individual dye. Among them, the ABFO-BPNN model constructed on the basis of combined dyes had the best prediction performance, with prediction correlation coefficient (RP2), root mean square error of the prediction set (RMSEP), and ratio of performance deviation (RPD) values of 0.9348, 0.3976, and 3.9695, respectively. The aim of this study was to develop a reliable evaluation model to achieve fast and accurate predictions of silage pH. Full article

Remnant explosive devices are a deadly nuisance to both military personnel and civilians. Traditional mine detection and clearing is dangerous, time-consuming, and expensive. And routine production and testing of explosives can create groundwater contamination issues. Remote detection methods could be rapidly deployed in vegetated areas containing explosives as they are known to cause stress in vegetation that is detectable with hyperspectral sensors. Hyperspectral imagery was employed in a mesocosm study comparing stress from a natural source (drought) to that of plants exposed to two different concentrations of Royal Demolition Explosive (RDX; 250 mg kg⁻¹, 500 mg kg⁻¹). Classification was accomplished with the machine learning algorithms Support Vector Machine (SVM), Random Forest (RF), and Least Discriminant Analysis (LDA). Leaf-level plant data assisted in validating plant stress induced by the presence of explosives and was detectable. Vegetation indices (VIs) have historically been used for dimension reduction due to computational limitations; however, we measured improvements in model precision, recall, and accuracy when using the complete range of available wavelengths. In fact, almost all models applied to spectral data outperformed their index counterparts. While challenges exist in scaling research efforts from the greenhouse to the field (i.e., weather, solar lighting conditions, altitude when imaging from a UAV, runoff containment, etc.), this experiment is promising for subsequent research efforts at greater scale and complexity aimed at detecting emerging contaminants. Full article

To achieve large-scale stitching of the hyperspectral remote sensing images obtained by unmanned aerial vehicles (UAVs) equipped with an acousto-optic tunable filter spectrometer, this study proposes a method based on a feature fusion strategy and a seam-finding strategy using hyperspectral image classification. In the feature extraction stage, SuperPoint deep features from images in different spectral segments of the data cube were extracted and fused. The feature depth matcher, LightGlue, was employed for feature matching. During the data cube fusion stage, unsupervised K-means spectral classification was performed separately on the two hyperspectral data cubes. Subsequently, grayscale transformations were applied to the classified images. A dynamic programming method, based on a grayscale loss function, was then used to identify seams in the transformed images. Finally, the identified splicing seam was applied across all bands to produce a unified hyperspectral data cube. The proposed method was applied to hyperspectral data cubes acquired at specific waypoints by UAVs using an acousto-optic tunable filter spectral imager. Experimental results demonstrated that the proposed method outperformed both single-spectral-segment feature extraction methods and stitching methods that rely on seam identification from a single spectral segment. The improvement was evident in both the spatial and spectral dimensions. Full article

In order to rapidly and nondestructively detect pesticide residues on tomato leaves, fluorescence spectroscopy and hyperspectral techniques were used to study the nondestructive detection of three different concentrations of benzyl-pyrazolyl esters on the surface of tomato leaves, respectively. In this study, fluorescence spectrum acquisition and hyperspectral imaging processing of tomato leaf samples with and without pesticides were conducted, and spectral data from regions of interest of hyperspectral images were extracted. The data in the spectral raw bands were optimized using convolutional smoothing (S-G), standard normal variable transformation (SNV), multiplicative scatter correction (MSC), and baseline calibration (baseline) algorithms, respectively. In order to improve the operating rate of discrimination, a continuous projection algorithm (SPA) was used to extract the characteristic wavelengths of the fluorescence spectra and hyperspectral data of pesticide residues, and algorithms such as the least-squares support vector machine (LSSVM) algorithm and least partial squares regression (PLSR) were used to build a quantitative model, while algorithms such as the convolutional neural network (BPNN) algorithm and decision tree algorithm (CART) were used to build a qualitative model. According to the results, R² of the model of hyperspectral data after SG-SNV preprocessing and PLSR modeling reached 0.9974, RMSEC reached 0.0221, and RMSEP reached 0.0565. R² of the model of fluorescence spectral data after SG-MSC preprocessing and SVM modeling reached 0.9986, RMSEC reached 0.2496, and RMSEP reached 0.4193. Qualitative analysis was established based on the characteristic wavelengths of hyper-spectrum and fluorescence spectrum extracted by the SPA algorithm, and the accuracy of the training sets of the optimal qualitative model reached 94.9% and 95.7%, respectively, and the accuracy of the test sets both reached 100%. After comparison, the quantitative model of data based on fluorescence spectrum for pesticide residue detection in tomato leaves proved to have a better effect, and the qualitative model showed higher accuracy in discrimination. Therefore, the fluorescence spectral and hyperspectral imaging techniques applied to tomato leaf pesticide detection enjoy a promising application prospect. Full article

Illumination is a crucial factor in hyperspectral imaging systems. In this respect, this work is focused on analyzing the influence of the light power source in acquiring hyperspectral images. To this end, a custom regulated power supply was designed and developed. This power supply was then integrated into a hyperspectral acquisition system, and several light stability measurements were conducted. Finally, several parameters related to the stability of the light produced by those systems were extracted using image analysis techniques, and a statistical comparison among the different power supplies was performed. Two commercial power supplies were also analyzed under the same experimental conditions and compared with the proposed power supply. The hyperspectral measurements were conducted using light transmission and reflectance. The results indicate that the proposed power supply performs better than or at least as well as commercial power supplies in terms of light stability. Additionally, this study shows the impact of power supply design on the stability and quality of hyperspectral illumination, especially concerning the signal-to-noise ratio (SNR) across different spectral bands. It is shown that optimizing the design of the power supply could improve light stability in hyperspectral imaging applications. Full article

Trees are indispensable to ecosystems, yet mortality rates have been increasing due to the abnormal changes in forest growth environments caused by frequent extreme weather events associated with global climate warming. Consequently, the need to monitor, assess, and predict tree mortality has become increasingly urgent to better address climate change and protect forest ecosystems. Over the past few decades, remote sensing has been widely applied to vegetation mortality observation due to its significant advantages. Here, we reviewed and analyzed the major research advancements in the application of remote sensing for tree mortality monitoring, using the Web of Science Core Collection database, covering the period from 1998 to the first half of 2024. We comprehensively summarized the use of different platforms (satellite and UAV) for data acquisition, the application of various sensors (multispectral, hyperspectral, and radar) as image data sources, the primary indicators, the classification models used in monitoring tree mortality, and the influence of tree mortality. Our findings indicated that satellite-based optical remote sensing data were the primary data source for tree mortality monitoring, accounting for 80% of existing studies. Time-series optical remote sensing data have emerged as a crucial direction for enhancing the accuracy of vegetation mortality monitoring. In recent years, studies utilizing airborne LiDAR have shown an increasing trend, accounting for 48% of UAV-based research. NDVI was the most commonly used remote sensing indicator, and most studies incorporated meteorological and climatic factors as environmental variables. Machine learning was increasingly favored for remote sensing data analysis, with Random Forest being the most widely used classification model. People are more focused on the impacts of tree mortality on water and carbon. Finally, we discussed the challenges in monitoring and evaluating tree mortality through remote sensing and offered perspectives for future developments. Full article

In recent years, convolutional neural network (CNN)-based and transformer-based approaches have made strides in improving the performance of hyperspectral image (HSI) classification tasks. However, misclassifications are unavoidable in the aforementioned methods, with a considerable number of these issues stemming from the overlapping embedding spaces among different classes. This overlap results in samples being allocated to adjacent categories, thus leading to inaccurate classifications. To mitigate these misclassification issues, we propose a novel discrete vector representation (DVR) strategy for enhancing the performance of HSI classifiers. DVR establishes a discrete vector quantification mechanism to capture and store distinct category representations in the codebook between the encoder and classification head. Specifically, DVR comprises three components: the Adaptive Module (AM), Discrete Vector Constraints Module (DVCM), and auxiliary classifier (AC). The AM aligns features derived from the backbone to the embedding space of the codebook. The DVCM employs category representations from the codebook to constrain encoded features for a rational feature distribution of distinct categories. To further enhance accuracy, the AC correlates discrete vectors with category information obtained from labels by penalizing these vectors and propagating gradients to the encoder. It is worth noting that DVR can be seamlessly integrated into HSI classifiers with diverse architectures to enhance their performance. Numerous experiments on four HSI benchmarks demonstrate that our DVR scheme improves the classifiers’ performance in terms of both quantitative metrics and visual quality of classification maps. We believe DVR can be applied to more models in the future to enhance their performance and provide inspiration for tasks such as sea ice detection and algal bloom prediction in the marine domain. Full article

Early detection of early-stage esophageal cancer (ECA) is crucial for timely intervention and improved treatment outcomes. Hyperspectral imaging (HSI) and artificial intelligence (AI) technologies offer promising avenues for enhancing diagnostic accuracy in this context. This study utilized a dataset comprising 3984 white light images (WLIs) and 3666 narrow-band images (NBIs). We employed the Yolov5 model, a state-of-the-art object detection algorithm, to predict early ECA based on the provided images. The dataset was divided into two subsets: RGB-WLIs and NBIs, and four distinct models were trained using these datasets. The experimental results revealed that the prediction performance of the training model was notably enhanced when using HSI compared to general NBI training. The HSI training model demonstrated an 8% improvement in accuracy, along with a 5–8% enhancement in precision and recall measures. Notably, the model trained with WLIs exhibited the most significant improvement. Integration of HSI with AI technologies improves the prediction performance for early ECA detection. This study underscores the potential of deep learning identification models to aid in medical detection research. Integrating these models with endoscopic diagnostic systems in healthcare settings could offer faster and more accurate results, thereby improving overall detection performance. Full article

With the increasing demand for wood in the wood market and the frequent trade of high-value wood, the accurate identification of wood varieties has become essential. This study employs two hyperspectral imaging systems—visible and near-infrared spectroscopy (VNIR) and short-wave infrared spectroscopy (SWIR)—in combination with a deep learning model to propose a method for wood species identification. Spectral data from wood samples were obtained through hyperspectral imaging technology, and classification was performed using a combination of convolutional neural networks (CNNs) and Transformer models. Multiple spectral preprocessing and feature extraction techniques were applied to enhance data quality and model performance. The experimental results show that the full-band modeling is significantly better than the feature-band modeling in terms of classification accuracy and robustness. Among them, the classification accuracy of SWIR reaches 100%, the number of model parameters is 1,286,228, the total size of the model is 4.93 MB, and the Floating Point Operations (FLOPs) is 1.29 M. Additionally, the Shapley Additive Explanation (SHAP) technique was utilized for model interpretability, revealing key spectral bands and feature regions that the model emphasizes during classification. Compared with other models, CNN-Transformer is more effective in capturing the key features. This method provides an efficient and reliable tool for the wood industry, particularly in wood processing and trade, offering broad application potential and significant economic benefits. Full article

Cotton, the most widely produced natural fiber, is integral to the textile industry and sustains the livelihoods of millions worldwide. However, its quality is frequently compromised by contamination, particularly from honeydew, a substance secreted by insects that leads to the formation of sticky fibers, thereby impeding textile processing. This study investigates ultraviolet (UV) hyperspectral imaging (230–380 nm) combined with multivariate data analysis to detect and quantify honeydew contaminations in real cotton samples. Reference cotton samples were sprayed multiple times with honey solutions to replicate the natural composition of honeydew. Comparisons were made with an alternative method where samples were soaked in sugar solutions of varying concentrations. Principal component analysis (PCA) and quadratic discriminant analysis (QDA) effectively differentiated and classified samples based on honey spraying times. Additionally, partial least squares regression (PLS-R) was utilized to predict the honeydew content for each pixel in hyperspectral images, achieving a cross-validation coefficient of determination R² = 0.75 and root mean square error of RMSE = 0.8 for the honey model. By employing a realistic spraying method that closely mimics natural contamination, this study refines sample preparation techniques for improved evaluation of honeydew levels. In conclusion, the integration of hyperspectral imaging with multivariate analysis represents a robust, non-destructive, and rapid approach for real-time detection of honeydew contamination in cotton, offering significant potential for industrial applications. Full article

Band selection is an effective way to reduce redundant information, while preserving the physical properties of hyperspectral images (HSI). However, most band selection methods merely consider the relevance and separability between pairs of bands and ignore those for different ground objects. To solve these issues, we propose a Unique Pixel extraction and Adaptive Neighbor Clustering (UPANC) band selection method in this theoretical study. First, in consideration of the characteristics of HSI data and tasks, unique pixels are obtained with a low-rank representation, where the importance of bands is analyzed from both spectral and spatial perspectives. Second, an adaptive neighbor clustering method is designed based on the unique pixels, which groups bands into several clusters through optimizing the graph structure under label smoothness. With support vector machines (SVM) as the classifier, the UPANC method achieved good performance, where the overall accuracy scores were 89.05%, 82.62%, and 92.07% on the Houston, IndianPines, and Pavia University datasets, respectively. The experimental results illustrated the advantages of the UPANC method, which could select optimal bands to enhance the performance in land cover observation. Full article

This study proposes an Additive Wavelet Transform (AWT)-based method to fuse Multispectral UAV (MS UAV, 5 cm resolution) and Sentinel-2 satellite imagery (10–20 m resolution), generating 5 cm resolution fused images with a focus on near-infrared and shortwave infrared bands to enhance the accuracy of mango canopy water content monitoring. The fused Sentinel-2 and MS UAV data were validated and calibrated using field-collected hyperspectral data to construct vegetation indices, which were then used with five machine learning (ML) models to estimate Fuel Moisture Content (FMC), Equivalent Water Thickness (EWT), and canopy water content (CWC). The results indicate that the addition of fused Sentinel-2 data significantly improved the estimation accuracy of all parameters compared to using MS UAV data alone, with the Genetic Algorithm Backpropagation Neural Network (GABP) model performing best (R² = 0.745, 0.859, and 0.702 for FMC, EWT, and CWC, respectively), achieving R² improvements of 0.066, 0.179, and 0.210. Slope, canopy coverage, and human activities were identified as key factors influencing the spatial variability of FMC, EWT, and CWC, with CWC being the most sensitive to environmental changes, providing a reliable representation of mango canopy water status. Full article

To evaluate the accuracy of Discrete Wavelet Transform (DWT) in monitoring the chlorophyll (CHL) content of maize canopies based on RGB images, a field experiment was conducted in 2023. Images of maize canopies during the jointing, tasseling, and grouting stages were captured using unmanned aerial vehicle (UAV) remote sensing to extract color, texture, and wavelet features and to construct a color and texture feature dataset and a fusion of wavelet, color, and texture feature datasets. Backpropagation neural network (BP), Stacked Ensemble Learning (SEL), and Gradient Boosting Decision Tree (GBDT) models were employed to develop CHL monitoring models for the maize canopy. The performance of these models was evaluated by comparing their predictions with measured CHL data. The results indicate that the dataset integrating wavelet features achieved higher monitoring accuracy compared to the color and texture feature dataset. Specifically, for the integrated dataset, the BP model achieved an R² value of 0.728, an RMSE of 3.911, and an NRMSE of 15.24%; the SEL model achieved an R² value of 0.792, an RMSE of 3.319, and an NRMSE of 15.34%; and the GBDT model achieved an R² value of 0.756, an RMSE of 3.730, and an NRMSE of 15.45%. Among these, the SEL model exhibited the highest monitoring accuracy. This study provides a fast and reliable method for monitoring maize growth in field conditions. Future research could incorporate cross-validation with hyperspectral and thermal infrared sensors to further enhance model reliability and expand its applicability. Full article