Rapid identification of chrysanthemum teas by computer vision and deep learning

Low-cost, immediate, general-purpose, and high-throughput (LIGHt) smartphone colorimetric screening assay for water-soluble protein

An efficient and rapid method for the detection of total soluble protein in tobacco leaves, utilizing a smartphone-based colorimetric approach has been developed. The proposed low-cost, immediate, general-purpose, and high-throughput (LIGHt) smartphone colorimetric screening assay integrates commercially available microplates, enabling on-site, high-throughput screening of tobacco leaf quality. The study involves preparing protein standard solutions and constructing standard curves using both spectrophotometric and smartphone-based methods. The LIGHt smartphone colorimetry yielded an average relative standard deviation of 10.6 %, a limit of detection of 2 μg/mL, and an average recovery of 93 %. The results demonstrated a comparable performance between intensities from the blue channel and the absorbance values in reflecting protein concentrations, validating the feasibility of utilizing smartphone colorimetry for protein concentration determination. Our approach demonstrates the potential for practical implementation in the field, providing a cost-effective and user-friendly solution for rapid quality assessment in the tobacco industry. The LIGHt smartphone colorimetry enhances quality control practices in the tobacco sector and offers a promising tool for on-site production quality testing in various industries, such as fruits and vegetables.

An instance-based deep transfer learning method for quality identification of Longjing tea from multiple geographical origins

For practitioners, it is very crucial to realize accurate and automatic vision-based quality identification of Longjing tea. Due to the high similarity between classes, the classification accuracy of traditional image processing combined with machine learning algorithm is not satisfactory. High-performance deep learning methods require large amounts of annotated data, but collecting and labeling massive amounts of data is very time consuming and monotonous. To gain as much useful knowledge as possible from related tasks, an instance-based deep transfer learning method for the quality identification of Longjing tea is proposed. The method mainly consists of two steps: (i) The MobileNet V2 model is trained using the hybrid training dataset containing all labeled samples from source and target domains. The trained MobileNet V2 model is used as a feature extractor, and (ii) the extracted features are input into the proposed multiclass TrAdaBoost algorithm for training and identification. Longjing tea images from three geographical origins, West Lake, Qiantang, and Yuezhou, are collected, and the tea from each geographical origin contains four grades. The Longjing tea from West Lake is regarded as the source domain, which contains more labeled samples. The Longjing tea from the other two geographical origins contains only limited labeled samples, which are regarded as the target domain. Comparative experimental results show that the method with the best performance is the MobileNet V2 feature extractor trained with a hybrid training dataset combined with multiclass TrAdaBoost with linear support vector machine (SVM). The overall Longjing tea quality identification accuracy is 93.6% and 91.5% on the two target domain datasets, respectively. The proposed method can achieve accurate quality identification of Longjing tea with limited samples. It can provide some heuristics for designing image-based tea quality identification systems.

Accurate Classification of Chunmee Tea Grade Using NIR Spectroscopy and Fuzzy Maximum Uncertainty Linear Discriminant Analysis

The grade of tea is closely related to tea quality, so the identification of tea grade is an important task. In order to improve the identification capability of the tea grade system, a fuzzy maximum uncertainty linear discriminant analysis (FMLDA) methodology was proposed based on maximum uncertainty linear discriminant analysis (MLDA). Based on FMLDA, a tea grade recognition system was established for the grade recognition of Chunmee tea. The process of this system is as follows: firstly, the near-infrared (NIR) spectra of Chunmee tea were collected using a Fourier transform NIR spectrometer. Next, the spectra were preprocessed using standard normal variables (SNV). Then, direct linear discriminant analysis (DLDA), maximum uncertainty linear discriminant analysis (MLDA), and FMLDA were used for feature extraction of the spectra, respectively. Finally, the k-nearest neighbor (KNN) classifier was applied to classify the spectra. The k in KNN and the fuzzy coefficient, m, were discussed in the experiment. The experimental results showed that when k = 1 and m = 2.7 or 2.8, the accuracy of the FMLDA could reach 98.15%, which was better than the other two feature extraction methods. Therefore, FMLDA combined with NIR technology is an effective method in the identification of tea grade.

Feature extraction of 3D Chinese rose model based on color and shape features

Flower classification is of great importance to the research fields of plants, food, and medicine. Due to more abundant information on three-dimensional (3D) flower models than two-dimensional 2D images, it makes the 3D models more suitable for flower classification tasks. In this study, a feature extraction and classification method were proposed based on the 3D models of Chinese roses. Firstly, the shape distribution method was used to extract the sharpness and contour features of 3D flower models, and the color features were obtained from the Red-Green-Blue (RGB) color space. Then, the RF-OOB method was employed to rank the extracted flower features. A shape descriptor based on the unique attributes of Chinese roses was constructed, χ² distance was adopted to measure the similarity between different Chinese roses. Experimental results show that the proposed method was effective for the retrieval and classification tasks of Chinese roses, and the average classification accuracy was approximately 87%, which can meet the basic retrieval requirements of 3D flower models. The proposed method promotes the classification of Chinese roses from 2D space to 3D space, which broadens the research method of flower classification.

Medicinal Chrysanthemum Detection under Complex Environments Using the MC-LCNN Model

Medicinal chrysanthemum detection is one of the desirable tasks of selective chrysanthemum harvesting robots. However, it is challenging to achieve accurate detection in real time under complex unstructured field environments. In this context, we propose a novel lightweight convolutional neural network for medicinal chrysanthemum detection (MC-LCNN). First, in the backbone and neck components, we employed the proposed residual structures MC-ResNetv1 and MC-ResNetv2 as the main network and embedded the custom feature extraction module and feature fusion module to guide the gradient flow. Moreover, across the network, we used a custom loss function to improve the precision of the proposed model. The results showed that under the NVIDIA Tesla V100 GPU environment, the inference speed could reach 109.28 FPS per image (416 × 416), and the detection precision (AP₅₀) could reach 93.06%. Not only that, we embedded the MC-LCNN model into the edge computing device NVIDIA Jetson TX2 for real-time object detection, adopting a CPU–GPU multithreaded pipeline design to improve the inference speed by 2FPS. This model could be further developed into a perception system for selective harvesting chrysanthemum robots in the future.

YOLO-VOLO-LS: A Novel Method for Variety Identification of Early Lettuce Seedlings

Accurate identification of crop varieties is an important aspect of smart agriculture, which is not only essential for the management of later crop differences, but also has a significant effect on unmanned operations in planting scenarios such as facility greenhouses. In this study, five kinds of lettuce under the cultivation conditions of greenhouses were used as the research object, and a classification model of lettuce varieties with multiple growth stages was established. First of all, we used the-state-of-the-art method VOLO-D1 to establish a variety classification model for the 7 growth stages of the entire growth process. The results found that the performance of the lettuce variety classification model in the SP stage needs to be improved, but the classification effect of the model at other stages is close to 100%; Secondly, based on the challenges of the SP stage dataset, we combined the advantages of the target detection mechanism and the target classification mechanism, innovatively proposed a new method of variety identification for the SP stage, called YOLO-VOLO-LS. Finally, we used this method to model and analyze the classification of lettuce varieties in the SP stage. The result shows that the method can achieve excellent results of 95.961, 93.452, 96.059, 96.014, 96.039 in Val-acc, Test-acc, Recall, Precision, F1-score, respectively. Therefore, the method proposed in this study has a certain reference value for the accurate identification of varieties in the early growth stage of crops.

A novel method based on machine vision system and deep learning to detect fraud in turmeric powder

Assessing the quality of food and spices is particularly important in ensuring proper human nutrition. The use of computer vision method as a non-destructive technique in measuring the quality of food and spices has always been taken into consideration by researchers. Due to the high nutritional value of turmeric among the spices as well as the fraudulent motives to gain economic profit from the selling of this product, its quality assessment is very important. The lack of marketability of grade 3 chickpeas (small and broken chickpeas) and their very low price have made them a good choice to be mixed with turmeric in powder form and sold in the market. In this study, an improved convolutional neural network (CNN) was used to classify turmeric powder images to detect fraud. CNN was improved through the use of gated pooling functions. We also show with a combined approach based on the integration of average pooling and max pooling that the accuracy and performance of the proposed CNN has increased. In this study, 6240 image samples were prepared in 13 categories (pure turmeric powder, chickpea powder, chickpea powder mixed with food coloring, 10, 20, 30, 40 and 50% fraud in turmeric). In the preprocessing step, unwanted parts of the image were removed. The data augmentation (DA) was used to reduce the overfitting problem on CNN. Also in this research, MLP, Fuzzy, SVM, GBT and EDT algorithms were used to compare the proposed CNN results with other classifiers. The results showed that prevention of the overfitting problem using gated pooling, the proposed CNN was able to grade the images of turmeric powder with 99.36% accuracy compared to other classifiers. The results of this study also showed that computer vision, especially when used with deep learning (DL), can be a valuable method in evaluating the quality and detecting fraud in turmeric powder.