Yiketai Ecological Technology Company and International AdvancedInstrument technology company cooperation, committed to providing "ecological-Agriculture-Comprehensive technical solution for health. The company is based on International AdvancedofSpecim hyperspectral imaging technology, combined with machine vision and automation research and development integration, provides users with diversified, customized, and automated online sorting solutions, achieving high-throughput rapid classification and real-time response of industrial assembly line products. The system can be combined with industrial assembly lines, conveyor belts, and intelligent sorting systems, based on powerful spectral recognition capabilities and flexible classification models, to output accurate recognition results in real time. Users do not need to perform a series of complex encoding and spectral image analysis work to obtain the final results, significantly reducing the threshold for hyperspectral imaging technology to enter the market application.

Key Features:

▌ SpectraScanSpectral imaging scanning platform technology, which can customize adaptation solutions according to user usage scenarios

Industrial grade push scan hyperspectral imaging instrument, optional400-1000nmThe900-1700nmThe2700-5300nmWaiting band

Classification model training software: Users can view sample data, train classification models, verify classification performance, and create applications

Online real-time sorting: a high-performance data processing unit that performs real-time calculations based on classification models and displays sorting results online

▌ SupportGigE，USB3.0, RS-232/485，CANInterface, compatible with downstream processes of industrial assembly lines, assisting intelligent sorting

Hyperspectral imaging technologyCETheFCCTheRoHS3Waiting for international mainstream certification

Application case:

1. Quality inspection of pistachios

The quality of nuts is usually evaluated based on the freshness of the product, the presence of defects, mites, and foreign substances. In the quality control of pistachios, defect and foreign substance detection is the most important link in production.

Previous methods such as HPLC, GC-MS, and IR-MS have the drawback of not being able to detect in real-time. In recent years, machine vision technology based on color imaging (RGB) has played an important role in food processing and quality control, but there are still shortcomings in the recognition of mites, molds, and other foreign objects. Italian researchers have applied hyperspectral imaging (HSI) technology to the quality control of pistachios, conducting qualitative and quantitative analysis of different physicochemical characteristics of samples to explore better strategies for food quality control.

The experimenters randomly selected different types of specific samples of pistachio mixture and divided 99 samples into 6 categories: edible pistachios (23), inedible pistachios (23), pistachio shells (13), pistachio skins (13), twigs (13), and seeds (14). As shown in Figure 1.1 on the left, the samples are divided into two groups: the training set (70%) and the validation set (30%). Firstly, the original spectra are preprocessed to highlight the feature differences of the six types of spectra.

Figure 1.1 Left: RGB image (a), training set (b), and validation set (c); Right: Average (a) and preprocessed (b) reflectance spectra of 6 types of samples within the SWIR range

The training dataset was trained using principal component analysis (PCA) and different classification models using multivariate classification methods. The training results are shown on the left in Figure 1.2, where the CART prediction graph is similar to the results of PCA kNN, achieving very good classification results and outperforming those obtained by PLS-DA and PCA-DA. Using different classification models to predict the validation dataset, the results are shown on the right in Figure 1.2. The models correctly identified stones, and the PLS-DA and PCA-DA prediction images are similar, showing the same misclassified areas. The CART model prediction image is characterized by more dispersed misclassified pixels. The PCA kNN prediction graph shows the bestAs a result, there are very few misclassified areas for a single category.

On the left (a, b, c, d) of Figure 1.2 are the training results; On the right (a, b, c, d) are the predicted results

The results indicate that the hyperspectral imaging technology based on SWIR band provides a good method for quality monitoring and control of pistachios, and has broad prospects in offline or real-time detection scenarios for industrial applications.

2. Non destructive quality testing and evaluation of herbal tea

With the increasing attention of consumers to health, the consumption of herbal tea is also gradually increasing. Herbal tea is a mixture of two or more plant species with the aim of improving taste and increasing health benefits. Like other foods or health products, quality control (QC) of herbal tea is crucial for ensuring food safety and health benefits.

Traditional quality control methods, such as high-performance liquid chromatography (HPLC) and mass spectrometry (MS), have high repeatability and accuracy, but are time-consuming and destructive. In the presence of multiple raw materials, each component needs to be tested separately, which is more time-consuming and labor-intensive. In this study, hyperspectral imaging technology was used as a fast and non-destructive method to evaluate and control the quality of herbal tea by combining traditional spectroscopy and digital imaging.

Researchers from Zwani University of Technology used SWIR hyperspectral imaging to capture hyperspectral images of raw materials (S.tortusosum and C. genistoides) purchased from Parceval Pty and five batches of herbal tea blends (Honeybush Celetium). In the absence of relevant chemical knowledge, a comprehensive and non-destructive distinction was made between the two raw materials, and principal component analysis (PCA) showed a 54.2% difference in chemical composition between the S. tortuosum and C. genistoides raw materials.

Figure 2.1 a) Two types of tea ingredients; b)PC1； c) Scatter plot of different pixel clusters of two types of tea (t1 vs t2)

In the rapid and non-destructive identification and component quantification of herbal tea, a PLS-DA model was developed based on the PCA calibration model, as shown on the left in Figure 2.2. The test set results were similar to PCA, observing two separate pixel clusters and obtaining a chemical composition difference of 52.8%. The corresponding Y-image is displayed next to the scatter plot, indicating that the pure raw material is 100% classified. On the right of Figure 2.2 are the RGB images and prediction results of the prediction set, which includes two external test set samples and five batches of herbal tea mixtures. The results confirm that the model can accurately predict the characteristics of tea mixture components. The predicted herbal tea mixture only contains two raw materials, S.tortuosum and C.genistoides, and quantitatively predicts that C.genistoides is the main component (content>97%), while S.tortuosum has a lower content (<3%). The quantitative results are close to the company's standard formula of C.genistoides=96% and S.tortuosum=4%. The experimental results indicate that hyperspectral imaging technology has great potential in the quality evaluation of food and drug products such as herbal tea blends.

Figure 2.2 Left: a) PLS-DA scatter plot and scores; b) The average spectral difference between S. tortuosum and C. genistoides. Right: Visualization of prediction set samples, a) RGB image; b) Class prediction based on PLS-DA model

Yiketai Ecological Technology Company provides customized spectral imaging solutions for food and traditional Chinese medicine sorting and quality control, seed and seedling sorting, industrial assembly line online sorting, waste recycling, production line quality control, machine vision applications, and other high-throughput application fields based on SpeatrAPP spectral imaging innovative application technology.

From left to right, they are peanut mold sorting, meat sorting (pork, beef, lamb), and apple sugar detection (provided by Yiketai Spectral Imaging Laboratory)

References:

[1] Bonifazi G , Capobianco G , Gasbarrone R , et al. Contaminant detection in pistachio nuts by different classification methods applied to short-wave infrared hyperspectral images[J]. Food Control, 2021, 130(26):108202.

[2] A M S , A W C , B I V A , et al. Non-destructive quality assessment of herbal tea blends using hyperspectral imaging - ScienceDirect[J]. Phytochemistry Letters, 2018, 24:94-101.