CHNSpec Technology （Zhejiang）Co.,Ltd Company News

In this study, a 400-1000nm hyperspectral camera can be used, and the products of Hangzhou Color Spectrum Technology Co., LTD FS13 conducts related research. The spectral range is 400-1000nm, and the wavelength resolution is better than 2.5nm, up to 1200 Two spectral channels. Acquisition speed up to 128FPS in the full spectrum, up to 3300Hz after band selection (multi-zone support Domain band selection). Chlorophyll plays an important role in plant photosynthesis, and its content is an important indicator of plant nutrient stress, photosynthetic capacity and growth status. The detection of plant chlorophyll content can be used to monitor plant growth and development, so as to scientifically guide cultivation and fertilization management, ensure good crop growth, improve crop quality and yield, which is of great significance for the practice of precision agriculture and forestry. The traditional chlorophyll content detection method is analytical chemistry method, that is, the leaves are collected in the laboratory, extracted by chemical solvent, and then the absorbance of the extracted liquid at two specific wavelengths is determined on the spectrophotometer, and the chlorophyll content is calculated according to the formula. This method has high measurement accuracy, but it is cumbersome, time-consuming and laborious, and it can not meet the requirements of rapid non-destructive testing in the field.   Visible near-infrared spectroscopy is a rapidly developed method of analysis and detection in recent years, which can make full use of spectral data at full spectrum or multi-wavelength for qualitative or quantitative analysis. Compared with the traditional analytical chemistry method, visible near-infrared spectroscopy has the characteristics of fast analysis, high efficiency, low cost, no damage, no pollution, etc., and has been widely used in many fields. In this paper, the vision-near-infrared spectral signals of plant leaves were obtained by means of transreflectance sampling, and the spectral data were preprocessed by smoothing, first-order differentiation and wavelet transform. Partial least square method (PLS) was used to establish the chlorophyll content and leaf absorption spectra of plant leaves. In this paper, a new method for the determination of chlorophyll content in plants by visible near-infrared spectroscopy was proposed. The reflectance sampling method is used to collect the spectrum of the blade, and the smoothing, differential and wavelet transform methods are used to preprocess the spectral data, which reduces the influence of non-target factors and improves the signal-to-noise ratio. Then, a quantitative analysis model of leaf chlorophyll content and leaf absorption spectrum was established by using partial least square method. The prediction accuracy of the model met the requirements of practical measurement applications. The results of this study showed that the application of vision-near-infrared spectroscopy to detect the chlorophyll content of leaves was feasible, which provided a basis for the rapid detection of the chlorophyll content of leaves, and also laid a foundation for the development of corresponding non-destructive testing instruments in the future.

In this study, a 900-1700nm hyperspectral camera was applied, and FS-15, the product of Hangzhou Color Spectrum Technology Co., LTD., could be used for related research. Short-wave near-infrared hyperspectral camera, the acquisition speed of the full spectrum up to 200FPS, is widely used in the composition identification, substance identification, machine vision, agricultural product quality, screen detection and other fields.   Anthocyanins are an important class of phenolic compounds in grape and wine, which mainly exist in the vacuoles of the cells in the 3-4 layers under the epidermis of grape berries. It is an important factor in determining the sensory quality of wine, and it is also the basis for the storage of red wine. The traditional chemical detection method will destroy the detection object, and it is difficult to achieve fast and large sample size detection. However, there are few studies on rapid detection of anthocyanins in wine grape fruits at home and abroad. In recent years, hyperspectral imaging technology as a non-destructive testing method has attracted wide attention, compared with the traditional near infrared spectroscopy technology, hyperspectral imaging technology shows its unique advantages. When using NIR spectroscopy, only one or several points of spectral information can be obtained each time, and there will be greater randomness and one-sidedness in the selection of the position and number of points. Hyperspectral image technology can obtain the image of the analyte, which not only provides more abundant information, but also provides a more reasonable and effective analysis method in spectral data processing. In the process of modeling using hyperspectral imaging technology combined with partial least squares method, with the deepening of the research on PLS method, it is found that better quantitative correction models may be obtained by screening characteristic wavelengths or wavelength intervals by specific methods.   In this experiment, the hyperspectral image of grape berries was obtained based on the near-infrared hyperspectral imaging system of 931 ~ 1700 nm. The continuous projection algorithm SPA was used to select the wavelength variables, and finally 20 spectral variables were selected from 236 wavelength points. Different modeling methods were used to establish the prediction model of anthocyanin content in grape peel. The results show that: (1) The continuous projection algorithm SPA can not only effectively select the characteristic spectral variables, simplify the correction model and shorten the correction time, but also improve the prediction accuracy of the model, which is an effective and practical method for the selection of spectral variables. (2) Among the four prediction models, PLS, SPA-MLR, SPA-BPNN and SPA-PLS, the SPA-PLS model has the best prediction effect and its prediction correlation coefficient R. And the predicted RMSEP were 0.9000 and 0.5506, respectively, maintaining a good forecast result. Therefore, the correlation between the spectral data of grape berries and the content of anthocyanins in grape skins is high. The near infrared hyperspectral imaging technology can effectively detect the content of anthocyanins in grape skins.

In this study, a 400-1000nm hyperspectral camera was applied, and FS13, a product of Hangzhou Color Spectrum Technology Co., LTD., could be used for related research. The spectral range is 400-1000nm, the wavelength resolution is better than 2.5nm, and up to 1200 spectral channels can be reached. The acquisition speed can reach 128FPS in the full spectrum, and the maximum after band selection is 3300Hz (support multi-region band selection). China's rice production accounts for more than 30% of the world's rice production, and "Meihe rice" in Jilin Province is a geographical indication product of China's japonica rice, and its production area is located in the world's golden grain production belt (45° N latitude). In practical life, there are many kinds of Meihe rice, and chemical methods such as Kjellod nitrogen determination and spectrophotometry are usually used to determine the protein content of different varieties of rice, but these traditional chemical methods are not only destructive to the sample itself, but also complicated steps and too long detection cycle. As a rapid non-destructive detection method, infrared spectroscopy has been widely used in the detection of the main components of rice (protein ≥, fat β, starch III, water), but it can only obtain the content of the components according to the spectral information, and can not achieve a more intuitive expression, that is, the visualization of the content. Hyperspectrum is a three-dimensional cube data including image information and spectral information. The obtained hyperspectral image contains both internal information of rice (internal physical structure and chemical composition information) and external information of rice (grain type, defects, etc.), which can make up for the lack of image that NIR can not quickly identify the spatial distribution of a certain substance. In this paper, rice of 3 varieties (Daohuahua, Akita Omachi and Jijing 60) from 4 producing areas in Meihe City, Jilin Province was selected as the research object. Hyperspectral imaging technology was used to detect the collected rice and obtain the average spectrum of the region of interest of the rice. In order to reduce the signal to noise ratio of the spectrum and obtain a relatively robust model, Three kinds of prediction models of rice protein content, including partial least square regression, principal component regression and error backpropagation neural network, were established by means of convolutional smoothing, mean centralization and multiple scattering correction. SPA was used to select the characteristic wavelength, establish the characteristic wavelength model, and transform the rice hyperspectral image into the protein content distribution map to realize the visualization of the protein content of rice from different varieties. The feasibility of visualization of protein content distribution in rice was studied by using hyperspectral imaging technology. A simplified and efficient PLSR protein content prediction model was obtained by MC spectral pretreatment method and the selection of SPA characteristic bands. Based on the quantitative model, the protein content distribution in rice of different varieties and different origin was visualized. Due to the similar shape of rice among the same varieties, it is difficult to distinguish the rice by ordinary RGB images. Imaging the protein content distribution can provide ideas for identifying the rice origin, and comparing the protein content distribution maps of rice among different varieties can provide evidence for the later breeding of rice varieties.

In this study, a 400-1000nm hyperspectral camera was applied, and FS13, a product of Hangzhou Color Spectrum Technology Co., LTD., could be used for related research. The spectral range is 400-1000nm, the wavelength resolution is better than 2.5nm, and up to 1200 spectral channels can be reached. The acquisition speed can reach 128FPS in the full spectrum, and the maximum after band selection is 3300Hz (support multi-region band selection).   Lettuce is rich in protein, carbohydrates, vitamins and other nutrients, and the planting area is wide. Nitrogen is one of the most important elements affecting the growth of lettuce. To establish a rapid, efficient and non-destructive method for nitrogen content detection of lettuce is convenient to guide the reasonable fertilization of lettuce. At present, there are few reports on the use of hyperspectral image technology to detect nitrogen content in lettuce leaves. In this study, hyperspectral image technology was applied to nondestructive detection of nitrogen content in lettuce leaves. By studying the effects of various spectral pretreatment methods on PLSB modeling, appropriate spectral pretreatment methods were selected for lettuce leaves, and sensitive wavelengths suitable for predicting nitrogen content in lettuce leaves were optimized. An attempt was made to establish the simplest and optimal prediction model of nitrogen content in lettuce leaves. This set of methods has not been reported, and it also provides a basis for the development of portable vegetable nutrient element detector, which has strong practical value.   The hyperspectral images of 60 lettuce leaves were collected by hyperspectral image technology, and the nitrogen content of the corresponding lettuce leaves was determined by AutoAnalyzer3 continuous flow analyzer. The average spectral data of 50×50 regions on the surface of the raw lettuce leaves was extracted by ENVI software. The extracted average spectral data were preprocessed (8 kinds of pretreatment methods). Finally, the original spectral data and 8 kinds of pretreatment spectral data were used as the input of PLSR to establish 9 prediction models for nitrogen content of lettuce. By comparing the results of these 9 prediction models, the optimal prediction model OSC+PLSR was selected, and the regression coefficient diagram of the OSC+PLSR model was analyzed. 13 sensitive wavelengths were selected, and then 13 sensitive wavelengths were taken as PLSR input. Finally, the prediction model of OSC+SW+PLSR lettuce nitrogen content was established. Compared with OSC+PLSR model, the prediction efficiency has been greatly improved, which can be used as an efficient, accurate and non-destructive new method for the prediction of nitrogen content in lettuce leaves, and can provide a reference for nitrogen nutrition diagnosis and economic and rational fertilization of lettuce.

In this study, a 400-1000nm hyperspectral camera can be used, and the products of Hangzhou Color Spectrum Technology Co., LTD FS13 conducts related research. The spectral range is 400-1000nm, and the wavelength resolution is better than 2.5nm, up to 1200 Two spectral channels. Acquisition speed up to 128FPS in the full spectrum, up to 3300Hz after band selection (multi-zone support Domain band selection).   With the promotion of potato staple grain strategy in China, potato related industry chain has been rapidly developed, and potato quality has become a hot issue. However, defects such as green skin and mechanical damage seriously affect the overall quantity of potatoes, especially the complex shape of green skin potatoes, defects are not easy to identify and increase the difficulty of detection. At the same time, if the content of solanin in green potato exceeds the edible standard, it will lead to food poisoning and cause food safety problems. Therefore, it is of great significance to study a fast and non-destructive detection method for potato deep processing and extension of potato industry chain.   Hyperspectral imaging technology has the advantages of wide band range, and can obtain the image and spectral information in the corresponding band range of the tested sample at the same time, so it has been widely used in rapid non-destructive testing of agricultural products. In order to solve the problem that the potato with light green skin is not easy to recognize under arbitrary position, the semi-transmission and reflection hyperspectral imaging techniques were used to compare and analyze, and the model recognition accuracy under different hyperspectral imaging methods was determined. The semi-transmitted hyperspectral and reflected hyperspectral images of potato samples were collected at any position, and detection models based on image information and spectral information were established respectively, and the recognition rates of different models were compared. Further establish image and spectrum fusion models or different imaging fusion models to improve model performance, and finally determine the optimal model. (1) The accuracy of image information recognition models with different hyperspectral imaging methods is compared. The recognition rate of isometric mapping combined with deep belief network model based on semi-transmitted image information is only 78.67%. The recognition rate of the maximum variance expansion combined with deep belief network model based on reflected image information is only 77.33%. The results showed that the accuracy of the detection of light green potato by single image information was not high. (2) The accuracy of spectral information recognition models with different hyperspectral imaging methods is compared. The recognition rate of local tangent space arrangement combined with deep belief network model based on semi-transmission spectrum information is the highest 93.33%. The recognition rate of local tangent spatial arrangement combined with deep belief network model based on reflectance spectral information is up to 90.67%. The results show that it is feasible to use single spectral information to detect light green potatoes, but the recognition rate needs to be further improved. (3) The influence of three multi-source information fusion methods on the recognition accuracy is compared. The accuracy of the three fusion models of semi-transmitted image and semi-transmitted spectrum, reflected image and reflection spectrum, semi-transmitted spectrum and reflection spectrum is higher than that of single image or spectral model, and the deep belief network fusion model of semi-transmitted spectrum and reflection spectrum is the best, and the recognition rate of correction set and test set is 100%. The results show that the fusion model of semi-transmission spectrum and reflection spectrum can realize the nondestructive testing of light green skin potato.

In this study, hyperspectral cameras of 400-1000nm band and 900-1700 nm were applied, and FS13 and FS15 products of Hangzhou Color Spectrum Technology Co., Ltd. could be used for related research. The spectral range is 400-1000nm, the wavelength resolution is better than 2.5nm, and up to 1200 spectral channels can be reached. The acquisition speed can reach 128FPS in the full spectrum, and the maximum after band selection is 3300Hz (support multi-region band selection). Meat mainly includes livestock and poultry and aquatic products, proteins, fatty acids, trace elements and other important energy substances needed by the human body are derived from meat. With the continuous improvement of living standards, people pay more attention to the quality of food and balanced nutrition in the diet, but some illegal businesses will mix some low-quality meat into high-quality meat, shoddy, especially in 2013 Europe's "horse meat wave", triggered people's extreme concern about meat adulteration. Meat adulteration detection methods include sensory evaluation, fluorescent PCR detection technology, electrophoresis analysis and enzyme-linked immunoassay technology, etc., but most of them require sample pretreatment, and the test operation is complicated and time-consuming, and it is difficult to achieve rapid real-time detection of large sample size in the field.   Most of the existing literature reports used single-band hyperspectral imaging technology to distinguish meat adulteration, but few used two bands for comparative analysis. In this experiment, high-quality defrosted mutton was selected as the adulterant, and duck meat with relatively low price was doped. Hyperspectral information of samples was collected in the two bands of visible near-infrared (400 ~ 1 000 nm) and short-wave near-infrared (900 ~ 1700 nm), and a quantitative model was established by selecting appropriate pretreatment methods. The optimal model was selected for image inversion, and a visualization method for rapid quantitative detection of mutton adulteration was proposed in order to provide data and technical support for the quantitative detection of mutton adulteration. (1) For the band of 400 ~ 1000 nm, the full-band PLS model established after normalization pretreatment has the highest accuracy; For the 900-1700 nm band, the full-band PLS model established after SNV pretreatment has the highest accuracy. By selecting the wavelength of the two spectral bands under the optimal pretreatment method, it is found that the collinearity between the selected wavelengths is minimal and representative on the basis of eliminating multicollinearity, which can further improve the accuracy and simplicity of the model.   (2) There is more information about groups related to meat composition in the 900-1700 nm band, which can better reflect the characteristics of meat, and may be more suitable for the identification of meat adulteration. In order to enlarge the comprehensiveness and applicability of the model, the experiment should be extended to the long wave near infrared spectrum (1 700 ~ 2500 nm). At the same time, the high-quality mutton and duck meat selected in the experiment were packaged as finished products in local supermarkets. Whether the subsequent model can be applied to the study of mutton adulteration under different environments (temperature, humidity, shape, etc.), different varieties, different qualities, different feeding methods and different freshness needs further verification and discussion.  

In this study, a 900-1700nm hyperspectral camera was applied, and FS-15, the product of Hangzhou Color Spectrum Technology Co., LTD., could be used for related research. Short-wave near-infrared hyperspectral camera, the acquisition speed of the full spectrum up to 200FPS, is widely used in the composition identification, substance identification, machine vision, agricultural product quality, screen detection and other fields.        Tomato is a berry crop with a unique flavor and rich in a variety of nutrients, including glutathione, vitamins, lycopene, beta-carotene and other bioactive ingredients, and has high food value. With the rapid development of the global economy, the demand for tomatoes and tomato processing products in the consumer market is increasing. The tomato has also become one of the most widely cultivated and consumed vegetable and fruit crops in the world. In addition, with the general improvement of people's living standards, the internal quality, appearance quality, storage and transportation quality and excellent flavor and taste of tomatoes have become more and more important to consumers, and China's tomato industry is also facing new challenges and opportunities. According to the survey, the maturity and storage quality of tomatoes are very important for the tomato industry, and the internal quality of cherry tomatoes, as well as excellent flavor and taste are more concerned by consumers. Based on the development and application of big data, automatic planting, mechanized picking and intelligent classification of tomatoes are realized to achieve increased production and efficiency of tomatoes. At present, there have been some researches on tomato quality detection based on spectrum at home and abroad, but in the existing tomato quality detection models, the extraction of effective spectral information is still a research difficulty, and the detection of tomato internal quality through appropriate non-destructive testing methods remains to be studied.     In the study of non-destructive detection of soluble solid content of cherry tomatoes based on hyperspectral imaging technology, 191 cherry tomatoes were selected as research objects, hyperspectral image data in the range of 865.11~1711.71 nm were collected, and the region of interest of cherry tomatoes hyperspectral image was segmsegmed by K-means algorithm. The average spectrum of this region was extracted as the original spectral data of cherry tomato. MA and MSC were used to preprocess the original spectral data, and the cherry tomato samples were divided into training sets and test sets based on KS algorithm. In order to improve the effectiveness of the information contained in the feature band, SPA algorithm and PCA algorithm were combined to perform principal component analysis on the spectral data, and then compared with PCA and miRF algorithms, a PLSR-based SSC detection model of cherry tomato was established, and the model was verified by the test set data. The results show that the detection accuracy of the model based on the principal component extracted by SPA-PCA is obviously optimized. From the detection results of the models, among the three models, SPA-PCA-PLSR model has the best detection effect, R, 0.9039. The detection effect of miRF-PLSR model was the second, RF was 0.8878. The fitting effect of PCA-PLSR model is the worst.

In this study, a 400-1000nm hyperspectral camera was applied, and FS13, a product of Hangzhou Color Spectrum Technology Co., LTD., could be used for related research. The spectral range is 400-1000nm, the wavelength resolution is better than 2.5nm, and up to 1200 spectral channels can be reached. The acquisition speed can reach 128FPS in the full spectrum, and the maximum after band selection is 3300Hz (support multi-region band selection). Blueberries, also known as blueberries, dark blue fruits, berries, so also known as blue berries, is one of the emerging small berries in China. Due to its unique health and nutritional value, it has many nutrients needed by the human body, excellent processing properties, etc., and has been paid attention to." The internal quality of blueberries has a great impact on the taste of blueberries, and is also one of the important indicators to evaluate the quality of blueberries. The traditional test method generally uses measuring device to detect the sugar content and hardness of blueberry. Because of the single detection index, long time consuming and destructive, these detection methods are difficult to be applied to the industrial detection of sugar content and hardness of fruit. Therefore, it is of great significance to develop a non-destructive and efficient method to detect the sugar content and hardness of blueberry based on internal quality.   Throughout the domestic and foreign research on fruit sugar content and hardness detection, it can be seen that the use of characteristic wavelength selection method can effectively reduce the dimension of hyperspectral image data, reduce the redundancy of spectral data, improve the calibration performance and detection efficiency of the model, and obtain good prediction results. It shows that these characteristic wavelength selection methods can be beneficial to realize online fruit detection. However, these studies are mainly aimed at the detection of single indicators, and multiple models need to be established to detect multiple indicators of fruit, which increases the complexity of data processing. Therefore, it is necessary to establish a model for multi-index detection to save time and improve the efficiency of online detection. In this study, hyperspectral imaging technology was used to propose a multi-stage feature wavelength selection method for detecting both sugar content and hardness of blueberries in hyperspectral images. Feature wavelength selection methods such as continuous projection algorithm or stepwise multiple linear regression were used successively to select the feature wavelengths that could reflect both sugar content and hardness of blueberries, and BP neural network model was used as the detection model. The sugar content and hardness of blueberry were predicted in order to realize rapid and non-destructive testing of the internal quality of blueberry, and to provide theoretical basis for the construction of online quality testing of blueberry.

In this study, a 400-1000nm hyperspectral camera can be used, and the products of Hangzhou Color Spectrum Technology Co., LTD In this study, a 400-1000nm hyperspectral camera can be used, and the products of Hangzhou Color Spectrum Technology Co., LTD FS13 conducts related research. The spectral range is 400-1000nm, and the wavelength resolution is better than 2.5nm, up to 1200 Two spectral channels. Acquisition speed up to 128FPS in the full spectrum, up to 3300Hz after band selection (multi-zone support Domain band selection). FS13 conducts related research. The spectral range is 400-1000nm, and the wavelength resolution is better than 2.5nm, up to 1200 Two spectral channels. Acquisition speed up to 128FPS in the full spectrum, up to 3300Hz after band selection (multi-zone support Domain band selection). The silkworm (Bombyx mori Linnaeus) is an economic insect that eats mulberry and spins silk, so it is also called the silkworm. Silkworms originated in ancient China and were gradually domesticated by the original silkworms inhabiting mulberry trees. As early as 5,000 years ago, the ancients had mastered the technology of planting mulberry and raising silkworms. In ancient times, sericulture made great contributions to the development of economy and culture. At present, mulberry silkworm industry promotes the development of rural economy, improves the living standard of farmers, and is one of the important sideline industries in agricultural production. In addition, the silkworm industry is in a leading position in the international market and plays an important role in world trade, creating a large number of foreign exchange reserves for our country. Therefore, the sustainable development of mulberry silkworm industry has extremely important economic value and significance. The traditional chemical detection technology needs to pretreat the tested samples, the operation process is complicated, and a lot of chemical reagents are consumed. The accuracy of enzymatic rapid detection technology is low, so it can only be used for primary screening. Spectral nondestructive testing technology is not representative because of one-sided information. Therefore, a fast, reliable and comprehensive nondestructive testing of mulberry leaves is sought.   The method of pesticide residue is of great significance in crop safety detection. Hyperspectral imaging technology is a new non-destructive testing technology combining imaging technology and spectrum technology, which has the advantages of no need to destroy the measured object, comprehensive information acquisition and high detection accuracy. In this paper, hyperspectral imaging technology combined with spectral processing and analysis methods were used to study the pesticide residues in mulberry leaves, not only to study whether there are pesticide residues in mulberry leaves and the identification of pesticide residues, but also to study the quantitative detection of chlorpyrifos pesticide residues in mulberry leaves. The research content of this paper provides technical support for sericulture industry and strong guarantee for sericulture farmers' income, and promotes the sustainable and in-depth development of sericulture industry, which has extremely important theoretical value and practical significance. In this paper, hyperspectral imaging technology combined with spectral processing and analysis methods was used to quantitatively detect the content of chlorpyrifos in mulberry leaves. Mulberry leaves with different chlorpyrifos residues were used as test objects to obtain hyperspectral images of mulberry leaves in the range of 390-1050nm by hyperspectral imager. ENVI software is used to determine the region of interest of the blade and calculate the average spectral data of the region. The correlation coefficients between the mean spectral data of mulberry leaf samples and the corresponding chemical values determined by gas chromatograph were calculated, and 5 waves were selected according to the waveform diagram of correlation coefficient and wavelength.   The wavelengths corresponding to peaks and troughs are used as characteristic wavelengths (561.25, 680.86, 706.58, 714.32, 724.66nm). Based on spectral data at characteristic wavelength, a quantitative detection model of mulberry leaf residues was established by using multiple linear regression and support vector regression. The correction set determination coefficient R² of the MLR prediction model is 0.730, the root mean square error RMSEC is 38.599, and the prediction set determination coefficient R is obtained. Is 0.637, and the root mean square error RMSEP is 47.146. The correction set determination coefficient R3 is 0.920, the root-mean-square error RMSEC is 21.073, the prediction set determination coefficient R3 is 0.874, and the root-mean-square error RMSEP is 27.719. Through comparative analysis: SVR prediction model has better performance than MLR prediction model, so vision-near-infrared hyperspectral imaging technology combined with SVR prediction model can be used to nondestructive detection of chlorpyrifos residues in mulberry leaves.

In this study, a 400-1000nm hyperspectral camera can be used, and FS13, a product of Hangzhou CHNSpec Technology Co., Ltd, LTD., can be used for related research. The spectral range is 400-1000nm, the wavelength resolution is better than 2.5nm, and up to 1200 spectral channels can be reached. The acquisition speed can reach 128FPS in the full spectrum, and the maximum after band selection is 3300Hz (support multi-region band selection). The main nutrients of the compound feed include water, ash, crude protein, calcium, total phosphorus and so on. The detection of the main nutrients of feed is an indispensable technical link in the production process and an important means to ensure the quality of feed products. The detection and analysis method of feed is the basis of its quality control. At present, the traditional chemical analysis method is generally used to determine the main nutrients of compound feed. The traditional method of determination is often time-consuming and labor-intensive, resulting in time lag, while the determination cost is high, and some even need to destroy the sample itself, which also has higher requirements for operators and laboratories. To explore a method for rapid detection of the main nutrients of compound feed, comprehensively promote and apply it to the actual test and analysis of feed enterprises, which has high social and economic benefits for improving the detection rate and promoting the development of the testing level of compound feed. Hyperspectral image detection is a high-tech set of computer vision and spectral detection, the use of hyperspectral image technology to obtain the sample information contains a large number of spectral information of the three-dimensional image block, it not only has a high spectral resolution, and the spectral information extracted from the image can be used to detect the internal quality of the sample. Therefore, hyperspectral image detection technology is more and more favored by scholars at home and abroad, and has been widely used in the quality detection of agricultural products, but the application research in compound feed is rarely reported. In this study, hyperspectral image technology was used to obtain the visible/near-infrared spectral information of experimental samples of compound feed, and a quantitative analysis model of the main nutrients in compound feed, such as moisture, ash, crude protein, calcium and total phosphorus, was established by using stoichiometric methods, and the model was verified, aiming to explore the feasibility of using hyperspectral imaging technology to detect the main nutrients in compound feed. It also provides a new idea and basis for the rapid detection of compound feed. In this study, hyperspectral image technology was used to establish quantitative analysis models of crude protein, crude ash, water, total phosphorus and calcium content in compound feed by means of abnormal sample removal, sample set division, optimal spectral pretreatment and characteristic band selection, combined with partial least square stoichiometry. The models were verified. The crude protein sample set divided by SPXY method and crude ash sample set divided by CG method, combined with the combination of AS, FD and SNV, the quantitative analysis model established in the characteristic band has the best effect. The correction set determination coefficient R& of the optimal crude protein model is 0.8373, the root-mean-square error RMSEC is 2.1327%, the relative analysis error RPDc is 2.4851, the validation set RV is 0.7778, RMSEP is 2.6155%, and RPDv is 2.1143. The optimal crude ash R&, RMSEC 1.0107%, RPDc 2.2064, RV 0.7758, RMSEP 1.0611% and RPDv 2.1204 were obtained. The quantitative analysis models of crude protein and crude ash both show good predictive performance and can be used for practical quantitative analysis. Water sample set divided by CG method combined with pretreatment of AS, OSC and Detrend has the best effect in the characteristic band. Its correction set RE is 0.6470, RMSEC is 1.8221%, RPD is 1.6849, validation set Ry is 0.6314, RMSEP is 1.6003%. RPDv is 1.9371, although the model can be used in practical quantitative analysis, its prediction accuracy still needs to be further optimized. The results of the quantitative analysis model obtained from the total phosphorus sample set divided by CG method combined with the pretreatment methods of AS, FD and SNV were optimal. The ratio of RS, RMSEC and RPD of the optimal model was 0.6038, 0.1656% and 1.5700, respectively. Validation sets R9, RMSEP and RPD/ are 0.4672, 0.1916% and 1.3570, respectively. The performance parameters of both the correction model and the validation model are poor, indicating that the model has poor predictive ability and cannot be used in actual quantitative analysis. After pretreatment of calcium sample set divided by CG method and combined with AS, OSC and Detrend method, the quantitative analysis model established in its characteristic band has the best effect, RB of the optimal model is 0.4784, and verification set R≈ is only 0.4406. The prediction effect of the model is poor, and it cannot be applied in practical analysis. The prediction accuracy of the crude protein optimal quantitative analysis model based on hyperspectral image technology is the best, and the prediction performance of the crude ash model is the second, and both can be used accurately in practical detection. The prediction accuracy of the water optimal quantitative analysis model should be improved. However, the optimal quantitative analysis model of total phosphorus and calcium has poor predictive performance and can not be used for practical detection.

In this study, a 400-1000nm hyperspectral camera was applied, and FS13, a product of Hangzhou CHNSpec Technology Co., Ltd, could be used for related research. The spectral range is 400-1000nm, the wavelength resolution is better than 2.5nm, and up to 1200 spectral channels can be reached. The acquisition speed can reach 128FPS in the full spectrum, and the maximum after band selection is 3300Hz (support multi-region band selection). Medicinal gelatin hollow hard capsule is a kind of special medicinal excipients, in which chromium content is an important test index stipulated by the national health standard. Capsules with excessive chromium content are commonly known as "toxic capsules" and are very toxic to the human body. At present, chromium content is determined by traditional chemical analysis method. The traditional chromium detection method is time-consuming, the equipment is expensive, the use of a large amount of nitric acid digestion is easy to cause secondary pollution, and the instrument operation needs professional personnel to complete. Therefore, the development of a convenient and rapid method for the rapid detection of chromium content in medicinal capsules has important application significance and market prospect.   Based on the feasibility of hyperspectral detection of heavy metals, this paper uses conventional atomic absorption spectrometry to compare the collected results of normal MEHGC and MEHGC with excessive chromium content, then collects two kinds of MehGC data with hyperspectral analysis, and uses principal component analysis (PCA) and partial least square method to analyze the hyperspectral data, and finally establishes the relevant model. To realize the qualitative detection of "poison capsules".   Since hyperspectral data is composed of multiple band images, each image can be regarded as a feature. If the hyperspectral data is dimensionally reduced, the original data will be changed to a new coordinate system to maximize the difference between the image data, and the result will be very different from the original image. This technique is very effective for enhancing information content, isolating noise and reducing data dimensions. The first 4 principal components obtained after PCA dimensionality reduction of hyperspectral images are shown in Figure 1. The advantage of hyperspectral images is that there is not only image information, but also spectral information. To obtain the spectral information, the region of interest is selected for each sample, and each region of interest has its spectral response curve. Due to the difference in color between the capsule cap and the capsule body, in order to eliminate the influence of color on the result, two regions of interest were selected for each capsule (one on the capsule cap and one on the capsule body). The regions of interest could be randomly selected on the hyperspectral image of the capsule, and the number of pixels in each region ranged from 2 to 6. The final spectral data for the region of interest is calculated as the average of all pixels in the region. The spectral curves of 4 different regions (capsules and caps of normal capsules and "toxic capsules" respectively) are shown in Figure 2. In the hyperspectral data of 450~900 nm, the spectral data of normal capsule and "toxic capsule" were obtained by selecting the region of interest, which was normalized first, and then the data dimension reduction and discriminant analysis were conducted by PLS-DA. When four PLS operators were selected as input features, the recognition rate of normal capsule and "toxic capsule" reached 100%. Specificity and sensitivity are also 100%; It can be seen that normal capsules and "toxic capsules" can be distinguished by PLS-DA discrimination method. Using hyperspectral image technology to detect "poison capsules" can greatly reduce the complexity of traditional methods.   In addition, to improve confidence, samples must be examined in a wider spectrum, such as fluorescence or ultraviolet. While qualitatively conducting the "poison capsule", it is also necessary to conduct quantitative research on it, which can consider making gelatin templates with different chromium content, find out the correlation model between the chromium content of the template and the spectral data, and use this model to predict the heavy metal chromium content of the unknown "poison capsule". In view of the subsequent impact of the "poison capsule" incident, samples are difficult to find, but in order to improve the effectiveness of the test, it is necessary to use a variety of capsule samples with chromium content.

In this study, a 400-1000nm hyperspectral camera can be used, and FS13, a product of Hangzhou CHNSpec Technology Co., Ltd, can be used for related research. The spectral range is 400-1000nm, and the wavelength resolution is better than 2.5nm, up to 1200 Two spectral channels. The acquisition speed can reach 128FPS in the full spectrum, and the maximum after band selection is 3300Hz (support multi-region band selection). Sorghum is one of the important food crops in China, because of its rich nutrients in the grain in the wine industry has a "good wine can not be separated from red grain" incisive judgment, the annual demand of up to 20 million t. At present, the main varieties of wine sorghum are Luzhou Red, Qinghuyang, Runuo No. 7 and other glutinous sorghum with high starch content. Because there are many kinds of sorghum and different producing areas, the contents of starch, protein, fat and tannin in the grain are very different, which leads to great differences in flavor, style, quality and yield of liquor. It can be seen that the accurate and efficient identification of sorghum varieties before the batch storage of sorghum raw materials has a very important guiding significance for the production of high-quality liquor, which can control the production process such as the time of bubbling grain, water consumption and steaming grain during the brewing process. The traditional identification methods mainly include manual empirical identification and biological sampling detection. The former is subject to subjective influence, low efficiency, and difficult to form a unified standard, while the latter is cumbersome, time-consuming and laborious. Both of them cannot meet the needs of modern liquor enterprises for identifying sorghum, so it is urgent to find a fast, accurate and simple sorghum variety classification and detection method. The objective of this study is to classify 11 sorghum varieties by combining spectral information and image information, and identify different sorghum varieties by optimizing hyperspectral technology and machine learning methods through comparison and external verification, so as to improve their accuracy and efficiency in application.   The original spectral curves of 550 samples of 11 categories of sorghum and the spectral curves after MSC pretreatment are shown in Figure 1. Each color represents a different category. In this paper, the identification of 11 varieties of sorghum was studied based on the combination of hyperspectral spectrum and image information. The hyperspectral images of sorghum were collected, 48 feature wavelengths were selected from the spectra after MSC pre-processing by SPA algorithm, and then the texture features of the images were extracted. SVM, PLS-DA and ELM classification models were established based on the texture features, full spectrum, feature spectrum and their combined image information, respectively. Finally, the data not involved in modeling was used for external verification. The results show that the SVM classification model based on the combination of feature spectrum and texture features has the best effect. The correct recognition rate of the test set is 95.3%, and the accuracy of the verification set is 91.8%. The combination of visible spectrum and image can effectively realize the rapid recognition of wine sorghum and improve the recognition accuracy of the model. This provides a theoretical basis for the detection of different brewing raw materials and the realization of brewing automation.