Hyperspectral Image Recovery Using a Color Camera for Detecting Colonies of Foodborne Pathogens on Agar Plate

윤승철,신태성,Gerald W. Heitschmidt,Kurt C. Lawrence,박보순,Gary Gamble 한국농업기계학회 2019 바이오시스템공학 Vol.44 No.3

Purpose Hyperspectral imaging often requires a special camera system to obtain spectral images. The cost for acquisition and process of hyperspectral images is usually much higher than color images. On the contrary, typical consumer-grade digital color cameras are much cheaper to obtain and process spatially high-resolution images than hyperspectral cameras. This paper is concerned with the development of a hyperspectral image recovery technique that can reconstruct hyperspectral images only from color images obtained by a digital color camera. Methods A sparse representation and least squares regression-based classification of foodborne pathogens on agar plates are presented. The target pathogen bacteria were the six representative non-O157 Shiga toxin-producing Escherichia coli (STEC) serogroups (O26, O45, O103, O111, O121, and O145). The wavelength range for the spectral recovery was from 400 to 1000 nm. Unlike many other studies using color charts with known and noise-free spectra for training their spectral recovery models, we directly used the hyperspectral and color images of real scenes for training the spectral recovery models. Both hyperspectral and color images were calibrated to percent reflectance values and then spatially registered. Two spectral recovery models including polynomial multivariate linear regression (MLR) and partial least squares regression (PLSR) were evaluated and compared by cross-validation and independent test. Results The spectral recovery results showed that the PLSR was more effective than the MLR. The classification accuracy measured with recovered spectra in an independent test set was about 5–10% less than the case of using the true hyperspectral images although the difference in maximum classification accuracy was only about 2%. Conclusion The results suggested the potential of a cost-effective color imaging system using hyperspectral image classification algorithms for differentiating pathogens in agar plates.

침엽수종 분류를 위한 초분광영상과 다중분광영상의 비교

조형갑 ( Hyung Gab Cho ),이규성 ( Kyu Sung Lee ) 대한원격탐사학회 2014 大韓遠隔探査學會誌 Vol.30 No.1

수종 간의 유사한 분광특성 때문에 기존의 다중분광영상을 이용한 수종분류는 한계가 있다. 본 연구에서는 경기도 광릉수목원에 분포하는 다섯 종류의 침엽수림을 분류하기 위하여 초분광영상과 다중분광영상의 적합성을 비교 분석하였다. 연구지역을 대상으로 두 종류의 항공 초분광영상(AISA, CASI)을 촬영하였으며, 비교 목적으로 초분광영상을 이용하여 모의 제작된 ETM+ 다중분광영상을 사용하였다. 영상분류에 사용된 영상은 초분광영상의 모든 밴드를 포함한 영상, PCA 및 MNF 기법으로 차원 축소된 영상, 그리고 분류등급의 분광분리도를 이용하여 소수의 밴드만을 추출한 영상이다. 또한 감독분류 과정에서 MLC, SAM, SVM 등 세 종류의 분류기를 적용하였다. 전체적으로 침엽수종의 분류에 있어서 초분광영상이 다중분광영상보다 높은 분류정확도를 제공하고 있다. 특히 중적외선 파장영역을 포함한 AISA-dual영상이 가장 좋은 분류결과를 보여주었다. 또한 많은 분광밴드를 가진 초분광영상을 MNF기법으로 차원 축소한 영상을 사용했을 때, 다른 영상보다 높은 분류결과가 나왔다. 감독 분류과정에서는 최대우도법(MLC)을 적용했을 때, 가장 높은 분류정확도를 얻었다. Multispectral image classification of individual tree species is often difficult because of the spectral similarity among species. In this study, we attempted to analyze the suitability of hyperspectral image to classify coniferous tree species. Several image sets and classification methods were applied and the classification results were compared with the ones from multispectral image. Two airborne hyperspectral images (AISA, CASI) were obtained over the study area in the Gwangneung National Forest. For the comparison, ETM+ multispectral image was simulated using hyperspectral images as to have lower spectral resolution. We also used the transformed hyperspectral data to reduce the data volume for the classification. Three supervised classification schemes (SAM, SVM, MLC) were applied to thirteen image sets. In overall, hyperspectral image provides higher accuracies than multispectral image to discriminate coniferous species. AISA-dual image, which include additional SWIR spectral bands, shows the best result as compared with other hyperspectral images that include only visible and NIR bands. Furthermore, MNF transformed hyperspectral image provided higher classification accuracies than the full-band and other band reduced data. Among three classifiers, MLC showed higher classification accuracy than SAM and SVM classifiers.

드론기반 시공간 초분광영상을 활용한 식생유무에 따른 하천 수심산정 기법 적용성 검토

권영화,김동수,유호준 한국수자원학회 2023 한국수자원학회논문집 Vol.56 No.4

Due to the revision of the River Act and the enactment of the Act on the Investigation, Planning, and Management of Water Resources, a regular bed change survey has become mandatory and a system is being prepared such that local governments can manage water resources in a planned manner. Since the topography of a bed cannot be measured directly, it is indirectly measured via contact-type depth measurements such as level survey or using an echo sounder, which features a low spatial resolution and does not allow continuous surveying owing to constraints in data acquisition. Therefore, a depth measurement method using remote sensing-LiDAR or hyperspectral imaging-has recently been developed, which allows a wider area survey than the contact-type method as it acquires hyperspectral images from a lightweight hyperspectral sensor mounted on a frequently operating drone and by applying the optimal bandwidth ratio search algorithm to estimate the depth. In the existing hyperspectral remote sensing technique, specific physical quantities are analyzed after matching the hyperspectral image acquired by the drone's path to the image of a surface unit. Previous studies focus primarily on the application of this technology to measure the bathymetry of sandy rivers, whereas bed materials are rarely evaluated. In this study, the existing hyperspectral image-based water depth estimation technique is applied to rivers with vegetation, whereas spatio-temporal hyperspectral imaging and cross-sectional hyperspectral imaging are performed for two cases in the same area before and after vegetation is removed. The result shows that the water depth estimation in the absence of vegetation is more accurate, and in the presence of vegetation, the water depth is estimated by recognizing the height of vegetation as the bottom. In addition, highly accurate water depth estimation is achieved not only in conventional cross-sectional hyperspectral imaging, but also in spatio-temporal hyperspectral imaging. As such, the possibility of monitoring bed fluctuations (water depth fluctuation) using spatio-temporal hyperspectral imaging is confirmed. 하천법 개정 및 수자원의 조사·계획 및 관리에 관한 법률 제정으로 하상변동조사를 정기적으로 실시하는 것이 의무화되었고, 지자체가 계획적으로 수자원을 관리할 수 있도록 제도가 마련되고 있다. 하상 지형은 직접 측량할 수 없기 때문에 수심 측량을 통해 간접적으로 이루어지고 있으며, 레벨측량이나 음향측심기를 활용한 접촉식 방법으로 이루어지고 있다. 접촉식 수심측량법은 자료수집이 제한적이기 때문에 공간해상도가 낮고 연속적인 측량이 불가능하다는 한계가 있어 최근에는 LiDAR나 초분광영상을 이용한 원격탐사를 이용한 수심측정 기술이 개발되고 있다. 개발된 초분광영상을 이용한 수심측정 기술은 접촉식 조사보다 넓은 지역을 조사할 수 있고, 잦은 빈도로 자료취득이 용이한 드론에 경량 초분광센서를 탑재하여 초분광영상을 취득하고, 최적 밴드비 탐색 알고리즘을 적용해 수심분포 산정이 가능하다. 기존의 초분광 원격탐사 기법은 드론의 경로비행으로 획득한 초분광영상을 면단위의 영상으로 정합한 후 특정 물리량에 대한 분석이 수행되었으며, 수심측정의 경우 모래하천을 대상으로 한 연구가 주를 이루었으며, 하상재료에 대한 평가는 이루어지지 않았었다. 본 연구에서는 기존의 초분광영상을 활용한 수심산정 기법을 식생이 있는 하천에 적용하고, 동일지역에서 식생을 제거한 후의 2가지 케이스에 대해서 시공간 초분광영상과 단면초분광영상에 모두 적용하였다. 연구결과, 식생이 없는 경우의 수심산정이 더 높은 정확도를 보였으며, 식생이 있는 경우에는 식생의 높이를 바닥으로 인식한 수심이 산정되었다. 또한, 기존의 단면초분광영상을 이용한 수심산정뿐만 아니라 시공간 초분광영상에서도 수심산정의 높은 정확도를 보여 시공간 초분광영상을 활용한 하상변동(수심변동) 추적의 가능성을 확인하였다.

Hyperspectral Imaging and Partial Least Square Discriminant Analysis for Geographical Origin Discrimination of White Rice

Mo, Changyeun,Lim, Jongguk,Kwon, Sung Won,Lim, Dong Kyu,Kim, Moon S.,Kim, Giyoung,Kang, Jungsook,Kwon, Kyung-Do,Cho, Byoung-Kwan Korean Society for Agricultural Machinery 2017 바이오시스템공학 Vol.42 No.4

Purpose: This study aims to propose a method for fast geographical origin discrimination between domestic and imported rice using a visible/near-infrared (VNIR) hyperspectral imaging technique. Methods: Hyperspectral reflectance images of South Korean and Chinese rice samples were obtained in the range of 400 nm to 1000 nm. Partial least square discriminant analysis (PLS-DA) models were developed and applied to the acquired images to determine the geographical origin of the rice samples. Results: The optimal pixel dimensions and spectral pretreatment conditions for the hyperspectral images were identified to improve the discrimination accuracy. The results revealed that the highest accuracy was achieved when the hyperspectral image's pixel dimension was $3.0mm{\times}3.0mm$. Furthermore, the geographical origin discrimination models achieved a discrimination accuracy of over 99.99% upon application of a first-order derivative, second-order derivative, maximum normalization, or baseline pretreatment. Conclusions: The results demonstrated that the VNIR hyperspectral imaging technique can be used to discriminate geographical origins of rice.

Hyperspectral Imaging and Partial Least Square Discriminant Analysis for Geographical Origin Discrimination of White Rice

모창연,임종국,권성원,임동규,김문수,김기영,강정숙,권경도,조병관 한국농업기계학회 2017 바이오시스템공학 Vol.42 No.4

Purpose: This study aims to propose a method for fast geographical origin discrimination between domestic and imported rice using a visible/near-infrared (VNIR) hyperspectral imaging technique. Methods: Hyperspectral reflectance images of South Korean and Chinese rice samples were obtained in the range of 400 nm to 1000 nm. Partial least square discriminant analysis (PLS-DA) models were developed and applied to the acquired images to determine the geographical origin of the rice samples. Results: The optimal pixel dimensions and spectral pretreatment conditions for the hyperspectral images were identified to improve the discrimination accuracy. The results revealed that the highest accuracy was achieved when the hyperspectral image’s pixel dimension was 3.0 mm × 3.0 mm. Furthermore, the geographical origin discrimination models achieved a discrimination accuracy of over 99.99% upon application of a first-order derivative, second-order derivative, maximum normalization, or baseline pretreatment. Conclusions: The results demonstrated that the VNIR hyperspectral imaging technique can be used to discriminate geographical origins of rice.

Prediction of Moisture Content of Grafted Cucumber Seedlings using Hyperspectral Image

( Sung Hyuk Jang ),( Ho Jun Lee ),( Hwa Dong Jang ),( Jae Su Lee ),( Yong Hyeon Kim ) 한국농업기계학회 2018 한국농업기계학회 학술발표논문집 Vol.23 No.1

This study was conducted to predict the moisture content of grafted cucumber seedlings using hyperspectral imaging system. The grafted cucumber seedlings were healed under a mixed red and blue LED. The hyperspectral images were obtained using a hyperspectral camera at shortwave infrared (SWIR) ranges. These images taken every 24 hours during the healing period of 144 hours, were preprocessed with brightness correction, smoothing, normalization and first-order derivative. The moisture content of grafted seedlings was determined just after acquiring the hyperspectral image. To make a prediction model and identify significant wavelengths, some statistical analysis including stepwise multiple linear regression (SMLR), the principal component regression (PCR) and the partial least squares regression (PLSR) between moisture content and spectral reflectance were performed. The correlation coefficient between spectral reflectance and the moisture content of grafted cucumber seedlings were high negative at the wavelengths of 1,400-1,500nm, 1,800-1,950nm and over 2,100nm. Most of prediction models included wavelengths in the ranges of 1,370-1,450nm that was known as the water absorption bands. However, a first-order derivative PLSR model was greatly affected by the change of spectral reflectance between 1,666nm and 1,672nm. Determination coefficients of the model predicting moisture content for calibration and validation dataset were 0.77 and 0.74, respectively. And the standard errors for calibration and prediction dataset were 1.17% and 0.96%, respectively. This result indicated that hyperspectral images could be applied to predict the moisture content of grafted cucumber seedling.

A Spectral-spatial Cooperative Noise-evaluation Method for Hyperspectral Imaging

Bing Zhou,Bingxuan Li,Xuan He,Hexiong Liu 한국광학회 2020 Current Optics and Photonics Vol.4 No.6

Hyperspectral images feature a relatively narrow band and are easily disturbed by noise. Accurate estimation of the types and parameters of noise in hyperspectral images can provide prior knowledge for subsequent image processing. Existing hyperspectral-noise estimation methods often pay more attention to the use of spectral information while ignoring the spatial information of hyperspectral images. To evaluate the noise in hyperspectral images more accurately, we have proposed a spectral-spatial cooperative noiseevaluation method. First, the feature of spatial information was extracted by Gabor-filter and K-means algorithms. Then, texture edges were extracted by the Otsu threshold algorithm, and homogeneous image blocks were automatically separated. After that, signal and noise values for each pixel in homogeneous blocks were split with a multiple-linear-regression model. By experiments with both simulated and real hyperspectral images, the proposed method was demonstrated to be effective and accurate, and the composition of the hyperspectral image was verified.

On-line fresh-cut lettuce quality measurement system using hyperspectral imaging

Mo, Changyeun,Kim, Giyoung,Kim, Moon S.,Lim, Jongguk,Lee, Kangjin,Lee, Wang-Hee,Cho, Byoung-Kwan Elsevier 2017 BIOSYSTEMS ENGINEERING Vol.156 No.-

<P>In this study, an online quality measurement system for detecting foreign substances on fresh-cut lettuce was developed using hyperspectral reflectance imaging. The online detection system with a single hyperspectral camera in the range of 400–1000 nm was able to detect contaminants on both surfaces of fresh-cut lettuce. Algorithms were developed for this system to detect contaminants such as slugs and worms. The optimal wavebands for discriminating between contaminants and sound lettuce as well as between contaminants and the conveyor belt were investigated using the one-way analysis of variance (ANOVA) method. The subtraction imaging (SI) algorithm to classify slugs resulted in a classification accuracy of 97.5%, sensitivity of 98.0%, and specificity of 97.0%. The ratio imaging (RI) algorithm to discriminate worms achieved classification accuracy, sensitivity, and specificity rates of 99.5%, 100.0%, and 99.0%, respectively. The overall results suggest that the online quality measurement system using hyperspectral reflectance imaging can potentially be used to simultaneously discriminate foreign substances on fresh-cut lettuces.</P> <P><B>Highlights</B></P> <P> <UL> <LI> We developed online fresh-cut lettuce quality measurement system. </LI> <LI> The online measurement system was capable to detect defects on both surfaces of fresh-cut lettuce. </LI> <LI> The multispectral imaging algorithms were developed to detect the foreign substances. </LI> <LI> The imaging algorithms for slug and worm achieved the accuracy of 97.5% and 99.5%, respectively. </LI> </UL> </P>

Detection of melamine in milk powders using near-infrared hyperspectral imaging combined with regression coefficient of partial least square regression model

Lim, Jongguk,Kim, Giyoung,Mo, Changyeun,Kim, Moon S.,Chao, Kuanglin,Qin, Jianwei,Fu, Xiaping,Baek, Insuck,Cho, Byoung-Kwan Elsevier 2016 Talanta Vol.151 No.-

<P><B>Abstract</B></P> <P>Illegal use of nitrogen-rich melamine (C<SUB>3</SUB>H<SUB>6</SUB>N<SUB>6</SUB>) to boost perceived protein content of food products such as milk, infant formula, frozen yogurt, pet food, biscuits, and coffee drinks has caused serious food safety problems. Conventional methods to detect melamine in foods, such as Enzyme-linked immunosorbent assay (ELISA), High-performance liquid chromatography (HPLC), and Gas chromatography–mass spectrometry (GC–MS), are sensitive but they are time-consuming, expensive, and labor-intensive. In this research, near-infrared (NIR) hyperspectral imaging technique combined with regression coefficient of partial least squares regression (PLSR) model was used to detect melamine particles in milk powders easily and quickly. NIR hyperspectral reflectance imaging data in the spectral range of 990–1700nm were acquired from melamine-milk powder mixture samples prepared at various concentrations ranging from 0.02% to 1%. PLSR models were developed to correlate the spectral data (independent variables) with melamine concentration (dependent variables) in melamine-milk powder mixture samples. PLSR models applying various pretreatment methods were used to reconstruct the two-dimensional PLS images. PLS images were converted to the binary images to detect the suspected melamine pixels in milk powder. As the melamine concentration was increased, the numbers of suspected melamine pixels of binary images were also increased. These results suggested that NIR hyperspectral imaging technique and the PLSR model can be regarded as an effective tool to detect melamine particles in milk powders.</P> <P><B>Highlights</B></P> <P> <UL> <LI> Melamine particles contained in milk powder were detected by NIR hyperspectral imaging. </LI> <LI> Regression coefficient values were used to reconstruct the PLS images. </LI> <LI> PLS images were used to discriminate the melamine pixels from milk powder pixels. </LI> <LI> Melamine particles at 200ppm in milk powder were confirmed without pretreatment. </LI> </UL> </P> <P><B>Graphical abstract</B></P> <P>[DISPLAY OMISSION]</P>

Applications of Hyperspectral Imaging and Convolutional Neural Networks (CNN) for Agricultural Products and Food Quality

( Hoonsoo Lee ),( Byoung-kwan Cho ) 한국농업기계학회 2019 한국농업기계학회 학술발표논문집 Vol.24 No.1

Hyperspectral imaging techniques have been used for decades to measure the quality and safety of food. Chemometrics methods such as principal components analysis (PCA), and partial least-squares (PLS) have mainly been employed for hyperspectral imaging analysis. However, the methods do not consider the relationship between the neighboring pixel information constituting an image. The objective of this study is to identify the applicability of the convolutional neural networks (CNN) method to hyperspectral images for the assessment of food quality. We compared the accuracy of the proposed CNN-based method with that of the chemometrics method. The results revealed that the 3-D CNN-based method provided competitive results for the assessment of food quality.