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

Advancing horizons in vegetable cultivation: a journey from ageold practices to high-tech greenhouse cultivation-a review

Vegetable cultivation stands as a pivotal element in the agricultural transformation illustrating a complex interplay between technological advancements, evolving environmental perspectives, and the growing global demand for food. This comprehensive review delves into the broad spectrum of developments in modern vegetable cultivation practices. Rooted in historical traditions, our exploration commences with conventional cultivation methods and traces the progression toward contemporary practices emphasizing the critical shifts that have refined techniques and outcomes. A significant focus is placed on the evolution of seed selection and quality assessment methods underlining the growing importance of seed treatments in enhancing both germination and plant growth. Transitioning from seeds to the soil, we investigate the transformative journey from traditional soil-based cultivation to the adoption of soilless cultures and the utilization of sustainable substrates like biochar and coir. The review also examines modern environmental controls highlighting the use of advanced greenhouse technologies and artificial intelligence in optimizing plant growth conditions. We underscore the increasing sophistication in water management strategies from advanced irrigation systems to intelligent moisture sensing. Additionally, this paper discusses the intricate aspects of precision fertilization, integrated pest management, and the expanding influence of plant growth regulators in vegetable cultivation. A special segment is dedicated to technological innovations, such as the integration of drones, robots, and state-of-the-art digital monitoring systems, in the cultivation process. While acknowledging these advancements, the review also realistically addresses the challenges and economic considerations involved in adopting cutting-edge technologies. In summary, this review not only provides a comprehensive guide to the current state of vegetable cultivation but also serves as a forward-looking reference emphasizing the critical role of continuous research and the anticipation of future developments in this field.

Integrating spectral and image information for prediction of cottonseed vitality

Cotton plays a significant role in people's lives, and cottonseeds serve as a vital assurance for successful cotton cultivation and production. Premium-quality cottonseeds can significantly enhance the germination rate of cottonseeds, resulting in increased cotton yields. The vitality of cottonseeds is a crucial metric that reflects the quality of the seeds. However, currently, the industry lacks a non-destructive method to directly assess cottonseed vitality without compromising the integrity of the seeds. To address this challenge, this study employed a hyperspectral imaging acquisition system to gather hyperspectral data on cottonseeds. This system enables the simultaneous collection of hyperspectral data from 25 cottonseeds. This study extracted spectral and image information from the hyperspectral data of cottonseeds to predict their vitality. SG, SNV, and MSC methods were utilized to preprocess the spectral data of cottonseeds. Following this preprocessing step, feature wavelength points of the cottonseeds were extracted using SPA and CARS algorithms. Subsequently, GLCM was employed to extract texture features from images corresponding to these feature wavelength points, including attributes such as Contrast, Correlation, Energy, and Entropy. Finally, the vitality of cottonseeds was predicted using PLSR, SVR, and a self-built 1D-CNN model. For spectral data analysis, the 1D-CNN model constructed after MSC+CARS preprocessing demonstrated the highest performance, achieving a test set correlation coefficient of 0.9214 and an RMSE of 0.7017. For image data analysis, the 1D-CNN model constructed after SG+CARS preprocessing outperformed the others, yielding a test set correlation coefficient of 0.8032 and an RMSE of 0.9683. In the case of fused spectral and image data, the 1D-CNN model built after SG+SPA preprocessing displayed the best performance, attaining a test set correlation coefficient of 0.9427 and an RMSE of 0.6872. These findings highlight the effectiveness of the 1D-CNN model and the fusion of spectral and image features for cottonseed vitality prediction. This research contributes significantly to the development of automated detection devices for assessing cottonseed vitality.

The CNN model aided the study of the clinical value hidden in the implant images

PURPOSE

This article aims to construct a new method to evaluate radiographic image identification results based on artificial intelligence, which can complement the limited vision of researchers when studying the effect of various factors on clinical implantation outcomes.

METHODS

We constructed a convolutional neural network (CNN) model using the clinical implant radiographic images. Moreover, we used gradient-weighted class activation mapping (Grad-CAM) to obtain thermal maps to present identification differences before performing statistical analyses. Subsequently, to verify whether these differences presented by the Grad-CAM algorithm would be of value to clinical practices, we measured the bone thickness around the identified sites. Finally, we analyzed the influence of the implant type on the implantation according to the measurement results.

RESULTS

The thermal maps showed that the sites with significant differences between Straumann BL and Bicon implants as identified by the CNN model were mainly the thread and neck area. (2) The heights of the mesial, distal, buccal, and lingual bone of the Bicon implant post-op were greater than those of Straumann BL (P < 0.05). (3) Between the first and second stages of surgery, the amount of bone thickness variation at the buccal and lingual sides of the Bicon implant platform was greater than that of the Straumann BL implant (P < 0.05).

CONCLUSION

According to the results of this study, we found that the identified-neck-area of the Bicon implant was placed deeper than the Straumann BL implant, and there was more bone resorption on the buccal and lingual sides at the Bicon implant platform between the first and second stages of surgery. In summary, this study proves that using the CNN classification model can identify differences that complement our limited vision.

Detecting the Minimum Limit on Wheat Stripe Rust in the Latent Period Using Proximal Remote Sensing Coupled with Duplex Real-Time PCR and Machine Learning

Wheat stripe rust (WSR) is an airborne disease that causes severe damage to wheat. The rapid and early detection of WSR is essential for the prevention and control of this disease. The minimum detection limit (MDL) is one of the most important characteristics of quantitative methods that can be used to determine the scope and applicability of a measurement technique. Three wheat cultivars were inoculated with Puccinia striiformis f.sp. tritici (Pst), and a spectrometer was used to collect the canopy hyperspectral data, and the Pst content was obtained via a duplex real-time polymerase chain reaction (PCR) during the latent period, respectively. The disease index (DI) and molecular disease index (MDI) were calculated. The regression tree algorithm was used to determine the MDL of the Pst based on hyperspectral feature parameters. The logistic, IBK, and random committee algorithms were used to construct the classification model based on the MDL. The results showed that when the MDL was 0.7, IBK had the best recognition accuracy. The optimal model, which used the spectral feature R_2nd.dv ((the second derivative of the original hyperspectral value)) and the modeling ratio 2:1, had an accuracy of 91.67% on the testing set and 90.67% on the 10-fold cross-validation. Thus, during the latent period, the MDL of Pst was determined using hyperspectral technology as 0.7.

Detecting different pesticide residues on Hami melon surface using hyperspectral imaging combined with 1D-CNN and information fusion

Efficient, rapid, and non-destructive detection of pesticide residues in fruits and vegetables is essential for food safety. The visible/near infrared (VNIR) and short-wave infrared (SWIR) hyperspectral imaging (HSI) systems were used to detect different types of pesticide residues on the surface of Hami melon. Taking four pesticides commonly used in Hami melon as the object, the effectiveness of single-band spectral range and information fusion in the classification of different pesticides was compared. The results showed that the classification effect of pesticide residues was better by using the spectral range after information fusion. Then, a custom multi-branch one-dimensional convolutional neural network (1D-CNN) model with the attention mechanism was proposed and compared with the traditional machine learning classification model K-nearest neighbor (KNN) algorithm and random forest (RF). The traditional machine learning classification model accuracy of both models was over 80.00%. However, the classification results using the proposed 1D-CNN were more satisfactory. After the full spectrum data was fused, it was input into the 1D-CNN model, and its accuracy, precision, recall, and F1-score value were 94.00%, 94.06%, 94.00%, and 0.9396, respectively. This study showed that both VNIR and SWIR hyperspectral imaging combined with a classification model could non-destructively detect different pesticide residues on the surface of Hami melon. The classification result using the SWIR spectrum was better than that using the VNIR spectrum, and the classification result using the information fusion spectrum was better than that using SWIR. This study can provide a valuable reference for the non-destructive detection of pesticide residues on the surface of other large, thick-skinned fruits.

Automatic detection of pesticide residues on the surface of lettuce leaves using images of feature wavelengths spectrum

The inappropriate application of pesticides to vegetable crops often results in environmental pollution, which seriously impacts the environment and human health. Given that current methods of pesticide residue detection are associated with issues such as low accuracy, high equipment cost, and complex flow, this study puts forward a new method for detecting pesticide residues on lettuce leaves. To establish this method, spectral analysis was used to determine the characteristic wavelength of pesticide residues (709 nm), machine vision equipment was improved, and a bandpass filter and light source of characteristic wavelength were installed to acquire leaf image information. Next, image preprocessing and feature information extraction were automatically implemented through programming. Several links were established for the training model so that the required feature information could be automatically extracted after the batch input of images. The pesticide residue detected using the chemical method was taken as the output and modeled, together with the input image information, using the convolutional neural network (CNN) algorithm. Furthermore, a prediction program was rewritten to generalize the input images during the prediction process and directly obtain the output pesticide residue. The experimental results revealed that when the detection device and method designed in this study were used to detect pesticide residues on lettuce leaves in a key state laboratory, the coefficient of determination of the equation reached 0.883, and the root mean square error (RMSE) was 0.134 mg/L, indicating high accuracy and that the proposed method integrated the advantages of spectrum detection and deep learning. According to comparison testing, the proposed method can meet Chinese national standards in terms of accuracy. Moreover, the improved machine vision equipment was less expensive, thus providing powerful support for the application and popularization of the proposed method.

Non-Destructive Detection of Soybean Pest Based on Hyperspectral Image and Attention-ResNet Meta-Learning Model

Soybean plays an important role in food, medicine, and industry. The quality inspection of soybean is essential for soybean yield and the agricultural economy. However, soybean pest is an important factor that seriously affects soybean yield, among which leguminivora glycinivorella matsumura is the most frequent pest. Aiming at the problem that the traditional detection methods have low accuracy and need a large number of samples to train the model, this paper proposed a detection method for leguminivora glycinivorella matsumura based on an A-ResNet (Attention-ResNet) meta-learning model. In this model, the ResNet network was combined with Attention to obtain the feature vectors that can better express the samples, so as to improve the performance of the model. As well, the classifier was designed as a multi-class support vector machine (SVM) to reduce over-fitting. Furthermore, in order to improve the training stability of the model and the prediction performance on the testing set, the traditional Batch Normalization was replaced by the Layer Normalization, and the Label Smooth method was used to punish the original loss. The experimental results showed that the accuracy of the A-ResNet meta-learning model reached 94.57 ± 0.19%, which can realize rapid and accurate nondestructive detection, and provides theoretical support for the intelligent detection of soybean pests.

Application of Near-Infrared Spectroscopy and Hyperspectral Imaging Combined with Machine Learning Algorithms for Quality Inspection of Grape: A Review

Grape is a fruit rich in various vitamins, and grape quality is increasingly highly concerned with by consumers. Traditional quality inspection methods are time-consuming, laborious and destructive. Near-infrared spectroscopy (NIRS) and hyperspectral imaging (HSI) are rapid, non-destructive and accurate techniques for quality inspection and safety assessment of agricultural products, which have great potential in recent years. The review summarized the applications and achievements of NIRS and HSI for the quality inspection of grapes for the last ten years. The review introduces basic principles, signal mode, data acquisition, analysis and processing of NIRS and HSI data. Qualitative and quantitative analysis were involved and compared, respectively, based on spectral features, image features and fusion data. The advantages, disadvantages and development trends of NIRS and HSI techniques in grape quality and safety inspection are summarized and discussed. The successful application of NIRS and HSI in grape quality inspection shows that many fruit inspection tasks could be assisted with NIRS and HSI.

Tobacco shred varieties classification using Multi-Scale-X-ResNet network and machine vision

The primary task in calculating the tobacco shred blending ratio is identifying the four tobacco shred types: expanded tobacco silk, cut stem, tobacco silk, and reconstituted tobacco shred. The classification precision directly affects the subsequent determination of tobacco shred components. However, the tobacco shred types, especially expanded tobacco silk and tobacco silk, have no apparent differences in macro-scale characteristics. The tobacco shreds have small size and irregular shape characteristics, creating significant challenges in their recognition and classification based on machine vision. This study provides a complete set of solutions aimed at this problem for screening tobacco shred samples, taking images, image preprocessing, establishing datasets, and identifying types. A block threshold binarization method is used for image preprocessing. Parameter setting and method performance are researched to obtain the maximum number of complete samples with acceptable execution time. ResNet50 is used as the primary classification and recognition network structure. By increasing the multi-scale structure and optimizing the number of blocks and loss function, a new tobacco shred image classification method is proposed based on the MS-X-ResNet (Multi-Scale-X-ResNet) network. Specifically, the MS-ResNet network is obtained by fusing the multi-scale Stage 3 low-dimensional and Stage 4 high-dimensional features to reduce the overfitting risk. The number of blocks in Stages 1-4 are adjusted from the original 3:4:6:3 to 3:4:N:3 (A-ResNet) and 3:3:N:3 (B-ResNet) to obtain the X-ResNet network, which improves the model's classification performance with lower complexity. The focal loss function is selected to reduce the impact of identification difficulty for different sample types on the network and improve its performance. The experimental results show that the final classification accuracy of the network on a tobacco shred dataset is 96.56%. The image recognition of a single tobacco shred requires 103 ms, achieving high classification accuracy and efficiency. The image preprocessing and deep learning algorithms for tobacco shred classification and identification proposed in this study provide a new implementation approach for the actual production and quality detection of tobacco and a new way for online real-time type identification of other agricultural products.

Spectral Preprocessing Combined with Deep Transfer Learning to Evaluate Chlorophyll Content in Cotton Leaves

Rapid determination of chlorophyll content is significant for evaluating cotton's nutritional and physiological status. Hyperspectral technology equipped with multivariate analysis methods has been widely used for chlorophyll content detection. However, the model developed on one batch or variety cannot produce the same effect for another due to variations, such as samples and measurement conditions. Considering that it is costly to establish models for each batch or variety, the feasibility of using spectral preprocessing combined with deep transfer learning for model transfer was explored. Seven different spectral preprocessing methods were discussed, and a self-designed convolutional neural network (CNN) was developed to build models and conduct transfer tasks by fine-tuning. The approach combined first-derivative (FD) and standard normal variate transformation (SNV) was chosen as the best pretreatment. For the dataset of the target domain, fine-tuned CNN based on spectra processed by FD + SNV outperformed conventional partial least squares (PLS) and squares-support vector machine regression (SVR). Although the performance of fine-tuned CNN with a smaller dataset was slightly lower, it was still better than conventional models and achieved satisfactory results. Ensemble preprocessing combined with deep transfer learning could be an effective approach to estimate the chlorophyll content between different cotton varieties, offering a new possibility for evaluating the nutritional status of cotton in the field.

Predicting Leaf Nitrogen Content in Cotton with UAV RGB Images

Rapid and accurate prediction of crop nitrogen content is of great significance for guiding precise fertilization. In this study, an unmanned aerial vehicle (UAV) digital camera was used to collect cotton canopy RGB images at 20 m height, and two cotton varieties and six nitrogen gradients were used to predict nitrogen content in the cotton canopy. After image-preprocessing, 46 hand features were extracted, and deep features were extracted by convolutional neural network (CNN). Partial least squares and Pearson were used for feature dimensionality reduction, respectively. Linear regression, support vector machine, and one-dimensional CNN regression models were constructed with manual features as input, and the deep features were used as inputs to construct a two-dimensional CNN regression model to achieve accurate prediction of cotton canopy nitrogen. It was verified that the manual feature and deep feature models constructed from UAV RGB images had good prediction effects. R2 = 0.80 and RMSE = 1.67 g kg−1 of the Xinluzao 45 optimal model, and R2 = 0.42 and RMSE = 3.13 g kg−1 of the Xinluzao 53 optimal model. The results show that the UAV RGB image and machine learning technology can be used to predict the nitrogen content of large-scale cotton, but due to insufficient data samples, the accuracy and stability of the prediction model still need to be improved.

Mapping the Corn Residue-Covered Types Using Multi-Scale Feature Fusion and Supervised Learning Method by Chinese GF-2 PMS Image

The management of crop residue covering is a vital part of conservation tillage, which protects black soil by reducing soil erosion and increasing soil organic carbon. Accurate and rapid classification of corn residue-covered types is significant for monitoring crop residue management. The remote sensing technology using high spatial resolution images is an effective means to classify the crop residue-covered areas quickly and objectively in the regional area. Unfortunately, the classification of crop residue-covered area is tricky because there is intra-object heterogeneity, as a two-edged sword of high resolution, and spectral confusion resulting from different straw mulching ways. Therefore, this study focuses on exploring the multi-scale feature fusion method and classification method to classify the corn residue-covered areas effectively and accurately using Chinese high-resolution GF-2 PMS images in the regional area. First, the multi-scale image features are built by compressing pixel domain details with the wavelet and principal component analysis (PCA), which has been verified to effectively alleviate intra-object heterogeneity of corn residue-covered areas on GF-2 PMS images. Second, the optimal image dataset (OID) is identified by comparing model accuracy based on the fusion of different features. Third, the 1D-CNN_CA method is proposed by combining one-dimensional convolutional neural networks (1D-CNN) and attention mechanisms, which are used to classify corn residue-covered areas based on the OID. Comparison of the naive Bayesian (NB), random forest (RF), support vector machine (SVM), and 1D-CNN methods indicate that the residue-covered areas can be classified effectively using the 1D-CNN-CA method with the highest accuracy (Kappa: 96.92% and overall accuracy (OA): 97.26%). Finally, the most appropriate machine learning model and the connected domain calibration method are combined to improve the visualization, which are further used to classify the corn residue-covered areas into three covering types. In addition, the study showed the superiority of multi-scale image features by comparing the contribution of the different image features in the classification of corn residue-covered areas.

Determination of adulteration in wheat flour using multi-grained cascade forest-related models coupled with the fusion information of hyperspectral imaging

Wheat flour (WF) is a common ingredient in staple foods. However, the presence of intentional or unintentional adulterants makes it difficult to guarantee WF quality. Multi-grained cascade forest (gcForest) model, a non-neural network deep learning structure, fused with image-spectra features from hyperspectral imaging (HSI) was employed for detecting adulterant type (peanut, walnut, or benzoyl peroxide) and the corresponding concentration (0.03%, 0.05%, 0.1%, 0.5%, 1%, and 2%). Based on the spectra of full wavelength and effective wavelength (EW) from hyperspectral images of WF samples, the gcForest-related models exhibited high performance (lowest ACC_P = 92.45%) and stability (lowest area under the curve = 0.9986). Furthermore, the fusion of the EW and the image features extracted by the symmetric all convolutional neural network (SACNN) was used to establish the gcForest-related models. The maximum accuracy improvement of the fusion feature model relative to the single spectral model and the image model was 2.45% and 44.37%, respectively. The results indicate that the gcForest-related model, combined with the image-spectra fusion feature of HSI, provides an effective tool for detection in food and agriculture.

Multichannel CNN Model for Biomedical Entity Reorganization

Biomedical researchers and biologists often search a large amount of literature to find the relationship between biological entities, such as drug-drug and compound-protein. With the proliferation of medical literature and the development of deep learning, the automatic extraction of biological entity interaction relationships from literature has shown great potential. The fundamental scope of this research is that the approach described in this research uses technologies like dynamic word vectors and multichannel convolution to learn a larger variety of relational expression semantics, allowing it to detect more entity connections. The extraction of biological entity relationships is the foundation for achieving intelligent medical care, which may increase the effectiveness of intelligent medical question answering and enhance the development of precision healthcare. In the past, deep learning methods have achieved specific results, but there are the following problems: the model uses static word vectors, which cannot distinguish polysemy; the weight of words is not considered, and the extraction effect of long sentences is poor; the integration of various models can improve the sample imbalance problem, the model is more complex. The purpose of this work is to create a global approach for eliminating different physical entity links, such that the model can effectively extract the interpretation of the expression relationship without having to develop characteristics manually. To this end, a deep multichannel CNN model (MC-CNN) based on the residual structure is proposed, generating dynamic word vectors through BERT (Bidirectional Encoder Representation from Transformers) to improve the accuracy of lexical semantic representation and uses multihead attention to capture the dependencies of long sentences and by designing the Ranking loss function to replace the multimodel ensemble to reduce the impact of sample imbalance. Tested on multiple datasets, the results show that the proposed method has good performance.

Identification of Rice Seed Varieties Based on Near-Infrared Hyperspectral Imaging Technology Combined with Deep Learning

Rice is one of the most important food crops in the world, and rice seed varieties are related to the yield and quality of rice. This study used near-infrared (NIR) hyperspectral technology with conventional machine learning methods (support vector machine (SVM), logistic regression (LR), and random forest (RF)) and deep learning methods (LeNet, GoogLeNet, and residual network (ResNet)) to establish variety identification models for five common types of rice seeds. Among the deep learning methods, the classification accuracies of most models were higher than 95%. This study further used the deep learning methods to establish variety identification models for 10 varieties of rice seeds without considering their types. Among them, the ResNet model had the best classification results. The classification accuracy on the test set was 86.08%. This study used the saliency map method to visualize each convolutional neural network (CNN) model to find the band region that contributed the most to the data. The results showed that the bands with the largest data contribution were mainly concentrated at approximately 1300-1400 nm and secondarily concentrated at approximately 1050-1250 nm. The overall results showed that NIR hyperspectral imaging technology combined with deep learning could effectively distinguish rice seeds of different varieties. This method provided an effective way to identify rice seed varieties in a quick and nondestructive manner.

Current State of Hyperspectral Remote Sensing for Early Plant Disease Detection: A Review

The development of hyperspectral remote sensing equipment, in recent years, has provided plant protection professionals with a new mechanism for assessing the phytosanitary state of crops. Semantically rich data coming from hyperspectral sensors are a prerequisite for the timely and rational implementation of plant protection measures. This review presents modern advances in early plant disease detection based on hyperspectral remote sensing. The review identifies current gaps in the methodologies of experiments. A further direction for experimental methodological development is indicated. A comparative study of the existing results is performed and a systematic table of different plants' disease detection by hyperspectral remote sensing is presented, including important wave bands and sensor model information.

Application of Laser-Induced Breakdown Spectroscopy Coupled With Spectral Matrix and Convolutional Neural Network for Identifying Geographical Origins of Gentiana rigescens Franch

Accurate geographical origin identification is of great significance to ensure the quality of traditional Chinese medicine (TCM). Laser-induced breakdown spectroscopy (LIBS) was applied to achieve the fast geographical origin identification of wild Gentiana rigescens Franch (G. rigescens Franch). However, LIBS spectra with too many variables could increase the training time of models and reduce the discrimination accuracy. In order to solve the problems, we proposed two methods. One was reducing the number of variables through two consecutive variable selections. The other was transforming the spectrum into spectral matrix by spectrum segmentation and recombination. Combined with convolutional neural network (CNN), both methods could improve the accuracy of discrimination. For the underground parts of G. rigescens Franch, the optimal accuracy in the prediction set for the two methods was 92.19 and 94.01%, respectively. For the aerial parts, the two corresponding accuracies were the same with the value of 94.01%. Saliency map was used to explain the rationality of discriminant analysis by CNN combined with spectral matrix. The first method could provide some support for LIBS portable instrument development. The second method could offer some reference for the discriminant analysis of LIBS spectra with too many variables by the end-to-end learning of CNN. The present results demonstrated that LIBS combined with CNN was an effective tool to quickly identify the geographical origin of G. rigescens Franch.

Citrus Huanglongbing Detection Based on Multi-Modal Feature Fusion Learning

Citrus Huanglongbing (HLB), also named citrus greening disease, occurs worldwide and is known as a citrus cancer without an effective treatment. The symptoms of HLB are similar to those of nutritional deficiency or other disease. The methods based on single-source information, such as RGB images or hyperspectral data, are not able to achieve great detection performance. In this study, a multi-modal feature fusion network, combining a RGB image network and hyperspectral band extraction network, was proposed to recognize HLB from four categories (HLB, suspected HLB, Zn-deficient, and healthy). Three contributions including a dimension-reduction scheme for hyperspectral data based on a soft attention mechanism, a feature fusion proposal based on a bilinear fusion method, and auxiliary classifiers to extract more useful information are introduced in this manuscript. The multi-modal feature fusion network can effectively classify the above four types of citrus leaves and is better than single-modal classifiers. In experiments, the highest accuracy of multi-modal network recognition was 97.89% when the amount of data was not very abundant (1,325 images of the four aforementioned types and 1,325 pieces of hyperspectral data), while the single-modal network with RGB images only achieved 87.98% recognition and the single-modal network using hyperspectral information only 89%. Results show that the proposed multi-modal network implementing the concept of multi-source information fusion provides a better way to detect citrus HLB and citrus deficiency.

Fast Identification of Soybean Seed Varieties Using Laser-Induced Breakdown Spectroscopy Combined With Convolutional Neural Network

Soybean seed purity is a critical factor in agricultural products, standardization of seed quality, and food processing. In this study, laser-induced breakdown spectroscopy (LIBS) as an effective technology was successfully used to identify ten varieties of soybean seeds. We improved the traditional sample preparation scheme for LIBS. Instead of grinding and squashing, we propose a time-efficient method by pressing soybean seeds into rubber sand filled with culture plates through a ruler to ensure a relatively uniform surface height. In our experimental scheme, three LIBS spectra were finally collected for each soybean seed. A majority vote based on three spectra was applied as the final decision judging the attribution of a single soybean seed. The results showed that the support vector machine (SVM) obtained the optimal identification accuracy of 90% in the prediction set. In addition, PCA-ResNet (propagation coefficient adaptive ResNet) and PCSA-ResNet (propagation coefficient synchronous adaptive ResNet) were designed based on typical ResNet structure by changing the way of self-adaption of propagation coefficients. Combined with a new form of input data called spectral matrix, PCSA-ResNet obtained the optimal performance with the discriminate accuracy of 91.75% in the prediction set. T-distributed stochastic neighbor embedding (t-SNE) was used to visualize the clustering process of the extracted features by PCSA-ResNet. For the interpretation of the good performance of PCSA-ResNet coupled with the spectral matrix, saliency maps were further applied to visually show the pixel positions of the spectral matrix that had a significant influence on the discrimination results, indicating that the content and proportion of elements in soybean seeds could reflect the variety differences.

Application of Hyperspectral Imaging for Maturity and Soluble Solids Content Determination of Strawberry With Deep Learning Approaches

Maturity degree and quality evaluation are important for strawberry harvest, trade, and consumption. Deep learning has been an efficient artificial intelligence tool for food and agro-products. Hyperspectral imaging coupled with deep learning was applied to determine the maturity degree and soluble solids content (SSC) of strawberries with four maturity degrees. Hyperspectral image of each strawberry was obtained and preprocessed, and the spectra were extracted from the images. One-dimension residual neural network (1D ResNet) and three-dimension (3D) ResNet were built using 1D spectra and 3D hyperspectral image as inputs for maturity degree evaluation. Good performances were obtained for maturity identification, with the classification accuracy over 84% for both 1D ResNet and 3D ResNet. The corresponding saliency maps showed that the pigments related wavelengths and image regions contributed more to the maturity identification. For SSC determination, 1D ResNet model was also built, with the determination of coefficient (R ²) over 0.55 of the training, validation, and testing sets. The saliency maps of 1D ResNet for the SSC determination were also explored. The overall results showed that deep learning could be used to identify strawberry maturity degree and determine SSC. More efforts were needed to explore the use of 3D deep learning methods for the SSC determination. The close results of 1D ResNet and 3D ResNet for classification indicated that more samples might be used to improve the performances of 3D ResNet. The results in this study would help to develop 1D and 3D deep learning models for fruit quality inspection and other researches using hyperspectral imaging, providing efficient analysis approaches of fruit quality inspection using hyperspectral imaging.