Application of Near-Infrared Hyperspectral Imaging for Detection of External Insect Infestations on Jujube Fruit

Study on Black Spot Disease Detection and Pathogenic Process Visualization on Winter Jujubes Using Hyperspectral Imaging System

In this work, the potential of a hyperspectral imaging (HSI) system for the detection of black spot disease on winter jujubes infected by Alternaria alternata during postharvest storage was investigated. The HSI images were acquired using two systems in the visible and near-infrared (Vis-NIR, 400-1000 nm) and short-wave infrared (SWIR, 1000-2000 nm) spectral regions. Meanwhile, the change of physical (peel color, weight loss) and chemical parameters (soluble solids content, chlorophyll) and the microstructure of winter jujubes during the pathogenic process were measured. The results showed the spectral reflectance of jujubes in both the Vis-NIR and SWIR wavelength ranges presented an overall downtrend during the infection. Partial least squares discriminant models (PLS-DA) based on the HSI spectra in Vis-NIR and SWIR regions of jujubes both gave satisfactory discrimination accuracy for the disease detection, with classification rates of over 92.31% and 91.03%, respectively. Principal component analysis (PCA) was carried out on the HSI images of jujubes to visualize their infected areas during the pathogenic process. The first principal component of the HSI spectra in the Vis-NIR region could highlight the diseased areas of the infected jujubes. Consequently, Vis-NIR HSI and NIR HSI techniques had the potential to detect the black spot disease on winter jujubes during the postharvest storage, and the Vis-NIR HSI spectral information could visualize the diseased areas of jujubes during the pathogenic process.

Study on Qualitative Impact Damage of Loquats Using Hyperspectral Technology Coupled with Texture Features

Bruising is one of the main problems in the post-harvest grading and processing of ‘Zaozhong 6’ loquats, reducing the economic value of loquats, and even food quality and safety problems are caused by it. Therefore, one of the main tasks in the post-harvest processing of loquats is to detect whether loquats are bruised, as well as the degree of bruising of loquats, to reduce the loss by proper treatment. An appropriate dimensionality reduction method can be used to reduce the redundancy of variables and improve the detection speed. The multispectral analysis method (MAM) has the advantage of accurate, rapid, and nondestructive detection, which was proposed to identify the different bruising degrees of loquats in this study. Firstly, the visible and near-infrared region (Vis–NIR, 400–1000 nm), the visible region (Vis, 400–780 nm), and the near-infrared region (NIR, 781–1000 nm) were analyzed using principal component analysis (PCA) to obtain the spectral regions and PC vectors, which could be used to effectively distinguish bruised loquats from normal loquats. Then, based on the selected second PC (PC2) score images, a morphological segmentation method (MSM) was proposed to distinguish bruised loquats from normal loquats. Furthermore, the weight coefficients of corresponding wavelength points of different degrees of bruising of loquats were analyzed, and the local extreme points and both sides of the interval were selected as the characteristic wavelength points for multi-spectral image processing. A gray level co-occurrence matrix (GLCM) was used to extract texture features and gray information from two-band ratio images K_782/999. Finally, the MAM was proposed to detect the degree of bruising of loquats, which included the spectral data of three characteristic wavelength points in the NIR region coupled with texture features of the two-band ratio images, and the classification accuracy was 91.3%. This study shows that the MAM can be used as an effective dimensionality reduction method. The method not only improves the effect of prediction but also simplifies the process of prediction and ensures the accuracy of classification. The MSM can be used for rapid detection of normal and bruised fruits, and the MAM can be used to classify the degree of bruising of bruised fruits. Consequently, the processed methods are effective and can be used for the rapid and nondestructive detection of the degree of bruising of fruit.

Imaging Spectroscopy and Machine Learning for Intelligent Determination of Potato and Sweet Potato Quality

Imaging spectroscopy has emerged as a reliable analytical method for effectively characterizing and quantifying quality attributes of agricultural products. By providing spectral information relevant to food quality properties, imaging spectroscopy has been demonstrated to be a potential method for rapid and non-destructive classification, authentication, and prediction of quality parameters of various categories of tubers, including potato and sweet potato. The imaging technique has demonstrated great capacities for gaining rapid information about tuber physical properties (such as texture, water binding capacity, and specific gravity), chemical components (such as protein, starch, and total anthocyanin), varietal authentication, and defect aspects. This paper emphasizes how recent developments in spectral imaging with machine learning have enhanced overall capabilities to evaluate tubers. The machine learning algorithms coupled with feature variable identification approaches have obtained acceptable results. This review briefly introduces imaging spectroscopy and machine learning, then provides examples and discussions of these techniques in tuber quality determinations, and presents the challenges and future prospects of the technology. This review will be of great significance to the study of tubers using spectral imaging technology.

Combining Multi-Dimensional Convolutional Neural Network (CNN) With Visualization Method for Detection of Aphis gossypii Glover Infection in Cotton Leaves Using Hyperspectral Imaging

Cotton is a significant economic crop. It is vulnerable to aphids (Aphis gossypii Glovers) during the growth period. Rapid and early detection has become an important means to deal with aphids in cotton. In this study, the visible/near-infrared (Vis/NIR) hyperspectral imaging system (376-1044 nm) and machine learning methods were used to identify aphid infection in cotton leaves. Both tall and short cotton plants (Lumianyan 24) were inoculated with aphids, and the corresponding plants without aphids were used as control. The hyperspectral images (HSIs) were acquired five times at an interval of 5 days. The healthy and infected leaves were used to establish the datasets, with each leaf as a sample. The spectra and RGB images of each cotton leaf were extracted from the hyperspectral images for one-dimensional (1D) and two-dimensional (2D) analysis. The hyperspectral images of each leaf were used for three-dimensional (3D) analysis. Convolutional Neural Networks (CNNs) were used for identification and compared with conventional machine learning methods. For the extracted spectra, 1D CNN had a fine classification performance, and the classification accuracy could reach 98%. For RGB images, 2D CNN had a better classification performance. For HSIs, 3D CNN performed moderately and performed better than 2D CNN. On the whole, CNN performed relatively better than conventional machine learning methods. In the process of 1D, 2D, and 3D CNN visualization, the important wavelength ranges were analyzed in 1D and 3D CNN visualization, and the importance of wavelength ranges and spatial regions were analyzed in 2D and 3D CNN visualization. The overall results in this study illustrated the feasibility of using hyperspectral imaging combined with multi-dimensional CNN to detect aphid infection in cotton leaves, providing a new alternative for pest infection detection in plants.

An Accuracy Improvement Method Based on Multi-Source Information Fusion and Deep Learning for TSSC and Water Content Nondestructive Detection in “Luogang” Orange

The objective of this study was to find an efficient method for measuring the total soluble solid content (TSSC) and water content of “Luogang” orange. Quick, accurate, and nondestructive detection tools (VIS/NIR spectroscopy, NIR spectroscopy, machine vision, and electronic nose), four data processing methods (Savitzky–Golay (SG), genetic algorithm (GA), multi-source information fusion (MIF), convolutional neural network (CNN) as the deep learning method, and a partial least squares regression (PLSR) modeling method) were compared and investigated. The results showed that the optimal TSSC detection method was based on VIS/NIR and machine vision data fusion and processing and modeling by SG + GA + CNN + PLSR. The R2 and RMSE of the TSSC detection results were 0.8580 and 0.4276, respectively. The optimal water content detection result was based on VIS/NIR data and processing and modeling by SG + GA + CNN + PLSR. The R2 and RMSE of the water content detection results were 0.7013 and 0.0063, respectively. This optimized method largely improved the internal quality detection accuracy of “Luogang” orange when compared to the data from a single detection tool with traditional data processing method, and provides a reference for the accuracy improvement of internal quality detection of other fruits.

Non-Destructive Technologies for Detecting Insect Infestation in Fruits and Vegetables under Postharvest Conditions: A Critical Review

In the last two decades, food scientists have attempted to develop new technologies that can improve the detection of insect infestation in fruits and vegetables under postharvest conditions using a multitude of non-destructive technologies. While consumers' expectations for higher nutritive and sensorial value of fresh produce has increased over time, they have also become more critical on using insecticides or synthetic chemicals to preserve food quality from insects' attacks or enhance the quality attributes of minimally processed fresh produce. In addition, the increasingly stringent quarantine measures by regulatory agencies for commercial import-export of fresh produce needs more reliable technologies for quickly detecting insect infestation in fruits and vegetables before their commercialization. For these reasons, the food industry investigates alternative and non-destructive means to improve food quality. Several studies have been conducted on the development of rapid, accurate, and reliable insect infestation monitoring systems to replace invasive and subjective methods that are often inefficient. There are still major limitations to the effective in-field, as well as postharvest on-line, monitoring applications. This review presents a general overview of current non-destructive techniques for the detection of insect damage in fruits and vegetables and discusses basic principles and applications. The paper also elaborates on the specific post-harvest fruit infestation detection methods, which include principles, protocols, specific application examples, merits, and limitations. The methods reviewed include those based on spectroscopy, imaging, acoustic sensing, and chemical interactions, with greater emphasis on the noninvasive methods. This review also discusses the current research gaps as well as the future research directions for non-destructive methods' application in the detection and classification of insect infestation in fruits and vegetables.

Non-destructive assessment of the myoglobin content of Tan sheep using hyperspectral imaging

This study aimed to develop simplified models for rapid and nondestructive monitoring myoglobin contents (DeoMb, MbO₂ and MetMb) during refrigerated storage of Tan sheep based on a hyperspectral imaging (HSI) system in the spectral range of 400-1000 nm. Partial least squares regression (PLSR) and least-squares support vector machines (LSSVM) were applied to correlate the spectral data with the reference values of myoglobin contents measured by a traditional method. In order to simplify the LSSVM models, competitive adaptive reweighted sampling (CARS) and Interval variable iterative space shrinkage approach (iVISSA) were used to select key wavelengths. The new CARS-LSSVM models of DeoMb and MbO₂ yielded good results, with R²p of 0.810 and 0.914, RMSEP of 1.127 and 2.598, respectively. The best model of MetMb was new iVISSA-CARS-LSSVM, with an R²p of 0.915 and RMSEP of 2.777. The overall results from this study indicated that it was feasible to predict myoglobin contents in Tan sheep using HSI.

Advanced Analysis of Roots and Tubers by Hyperspectral Techniques

Hyperspectral techniques in terms of spectroscopy and hyperspectral imaging have become reliable analytical tools to effectively describe quality attributes of roots and tubers (such as potato, sweet potato, cassava, yam, taro, and sugar beet). In addition to the ability for obtaining rapid information about food external or internal defects including sprout, bruise, and hollow heart, and identifying different grades of food quality, such techniques have also been implemented to determine physical properties (such as color, texture, and specific gravity) and chemical constituents (such as protein, vitamins, and carotenoids) in root and tuber products with avoidance of extensive sample preparation. Developments of related quality evaluation systems based on hyperspectral data that determine food quality parameters would bring about economic and technical values to the food industry. Consequently, a comprehensive review of hyperspectral literature is carried out in this chapter. The spectral data acquired, the multivariate statistical methods used, and the main breakthroughs of recent studies on quality determinations of root and tuber products are discussed and summarized. The conclusion elaborates the promise of how hyperspectral techniques can be applied for non-invasive and rapid evaluations of tuber quality properties.

Near-infrared hyperspectral imaging for detection and quantification of azodicarbonamide in flour

BACKGROUND

The present study aimed to establish a method for the detection and quantification of azodicarbonamide (ADC) in flour using hyperspectral imaging technology. Hyperspectral images of pure flour, pure ADC and flour-ADC mixtures with different concentrations of ADC were collected. F-values of one-way analysis of variance for all possible wavebands within the spectra of the flour and ADC were calculated, and the maximum value indicated that the two wavebands have more significant differences, i.e. the optimal two wavebands. Threshold segmentation was used for band ratio images of two wavebands to create a binary image. This allowed visual identification of ADC-rich pixels in the mixtures.

RESULTS

The two wavebands with the largest difference between flour and ADC were 2039 nm and 1892 nm. Using the binary image construction method, different concentrations of ADC in flour were identified. The minimum detected concentration was 0.2 g kg^-1 . In the mixtures, the number of ADC-rich pixels detected had a good linear relationship with the ADC concentrations, with a correlation coefficient of 0.9845.

CONCLUSION

This study indicated that the band ratio algorithm combination with threshold segmentation for hyperspectral images provides a non-destructive method for detecting and quantifying of ADC in flour. © 2017 Society of Chemical Industry.