• Journal of The Korean Society of Grassland and Forage Science
• /
• v.29 no.3
• /
• pp.253-262
• /
• 2009

The present study was conducted to develop a rapid and accurate method of evaluating chemical composition of total mixed ration (TMR) for dairy cows using near infrared reflectance spectroscopy (NIRS). A total of 253 TMR samples were collected from TMR manufacturers and dairy farms in Korea. Prior to NIR analysis, TMR samples were dried at $65^{\circ}C$ for 48 hour and then ground to 2 mm size. The samples were scanned at 2 nm interval over the wavelength range of 400-2500 nm on a FOSS-NIR Systems Model 6500. The values obtained by NIR analysis and conventional chemical methods were compared. Generally, the relationship between chemical analysis and NIR analysis was linear: $R^2$ and standard error of calibration (SEC) were 0.701 (SEC 0.407), 0.965 (SEC 0.315), 0.796 (SEC 0.406), 0.889 (SEC 0.987), 0.894 (SEC 0.311), 0.933 (SEC 0.885) and 0.889 (SEC 1.490) for moisture, crude protein, ether extract, crude fiber, crude ash, acid detergent fiber (ADF) and neutral detergent fiber (NDF), respectively. In addition, the standard error of prediction (SEP) value was 0.371, 0.290, 0.321, 0.380, 0.960, 0.859 and 1.446 for moisture, crude protein, ether extract, crude fiber, crude ash, ADF and NDF, respectively. The results of the present study showed that the NIR analysis for unknown TMR samples would be relatively accurate. Use of the developed NIR calibration curve can obtain fast and reliable data on chemical composition of TMR. Collection and analysis of more TMR samples will increase accuracy and precision of NIR analysis to TMR.

• Korean Journal of Food Science and Technology
• /
• v.40 no.6
• /
• pp.603-610
• /
• 2008

The principal objective of current study was to evaluate the potential of near infrared reflectance spectroscopy (NIRS) as a non-destructive method for the prediction of the amylose and protein contents of un-hulled and brown rice in broad-based calibration models. The average amylose and protein content of 75 rice accessions were 20.3% and 7.1%, respectively. Additionally, the range of amylose and protein content were 16.6-24.5% and 3.8-9.3%, respectively. In total, 79 rice germplasms representing a wide range of chemical characteristics, variable physical properties, and origins were scanned via NIRS for calibration and validation equations. The un-hulled and brown rice samples evidenced distinctly different patterns in a wavelength range from 1,440 nm to 2,400 nm in the original NIR spectra. The optimal performance calibration model could be obtained by MPLS (modified partial least squares) using the first derivative method (1:4:4:1) for un-hulled rice and the second derivative method (2:4:4:1) for brown rice. The correlation coefficients $(r^2)$ and standard error of calibration (SEC) of protein and amylose contents for the un-hulled rice were 0.86, 2.48, and 0.84, 1.13, respectively. The $r^2$ and SEC of protein and amylose content for brown rice were 0.95, 1.09 and 0.94, 0.42, respectively. The results of this study suggest that the NIRS technique could be utilized as a routine procedure for the quantification of protein and amylose contents in large accessions of un-hulled rice germplasms.

• KOREAN JOURNAL OF CROP SCIENCE
• /
• v.53 no.2
• /
• pp.217-223
• /
• 2008

Amylose and protein contents are important traits determining the edible quality of rice, especially in East Asian countries. Near-Infrared Reflectance Spectroscopy (NIRS) has become a powerful tool for rapid and nondestructive quantification of natural compounds in agricultural products. To test the practically of using NIRS for estimation of brown rice amylose and protein contents, the spectral reflectances ($400{\sim}2500\;nm$) of total 9,483 accessions of rice germplasm in Rural development Administration (RDA) Genebank ere obtained and compared to chemically determined amylose and protein content. The protein content of tested 119 accessions ranged from 6.5 to 8.0% and 25 accessions exhibited protein contents between 8.5 to 9.5%. In case of amylose content, all tested accessions ranged from 18.1 to 21.7% and the grade from 18.1 to 19.9% includes most number of accessions as 152 and 4 accessions exhibited amylose content between 20.5 to 21.7%. The optimal performance calibration model could be obtained from original spectra of brown rice using MPLS (Modified Partial Least Squares) with the correlation coefficients ($r_2$) for amylose and protein content were 0.865 and 0.786, respectively. The standard errors of calibration (SEC) exhibited good statistic values: 2.078 and 0.442 for amylose and protein contents, respectively. All these results suggest that NIR spectroscopy may serve as reputable and rapid method for quantification of brown rice protein and amylose contents in large numbers of rice germplasm.

• Korean Journal of Remote Sensing
• /
• v.39 no.6_1
• /
• pp.1211-1224
• /
• 2023

Clouds cause many difficult problems in observing land surface phenomena using optical satellites, such as national land observation, disaster response, and change detection. In addition, the presence of clouds affects not only the image processing stage but also the final data quality, so it is necessary to identify and remove them. Therefore, in this study, we developed a new cloud detection technique that automatically performs a series of processes to search and extract the pixels closest to the spectral pattern of clouds in satellite images, select the optimal threshold, and produce a cloud mask based on the threshold. The cloud detection technique largely consists of three steps. In the first step, the process of converting the Digital Number (DN) unit image into top-of-atmosphere reflectance units was performed. In the second step, preprocessing such as Hue-Value-Saturation (HSV) transformation, triangle thresholding, and maximum likelihood classification was applied using the top of the atmosphere reflectance image, and the threshold for generating the initial cloud mask was determined for each image. In the third post-processing step, the noise included in the initial cloud mask created was removed and the cloud boundaries and interior were improved. As experimental data for cloud detection, CAS500-1 L2G images acquired in the Korean Peninsula from April to November, which show the diversity of spatial and seasonal distribution of clouds, were used. To verify the performance of the proposed method, the results generated by a simple thresholding method were compared. As a result of the experiment, compared to the existing method, the proposed method was able to detect clouds more accurately by considering the radiometric characteristics of each image through the preprocessing process. In addition, the results showed that the influence of bright objects (panel roofs, concrete roads, sand, etc.) other than cloud objects was minimized. The proposed method showed more than 30% improved results(F1-score) compared to the existing method but showed limitations in certain images containing snow.

• Korean Journal of Remote Sensing
• /
• v.33 no.5_1
• /
• pp.559-570
• /
• 2017

A partial least squares regression (PLSR) model was developed to map the internal soluble solids content (SSC) of apples using a ground-based hyperspectral scanner that could simultaneously acquire outdoor data and capture images of large quantities of apples. We evaluated the applicability of various preprocessing techniques to construct an optimal prediction model and calculated the optimal band through a variable importance in projection (VIP)score. From the 515 bands of hyperspectral images extracted at wavelengths of 360-1019 nm, 70 reflectance spectra of apples were extracted, and the SSC ($^{\circ}Brix$) was measured using a digital photometer. The optimal prediction model wasselected considering the root-mean-square error of cross-validation (RMSECV), root-mean-square error of prediction (RMSEP) and coefficient of determination of prediction $r_p^2$. As a result, multiplicative scatter correction (MSC)-based preprocessing methods were better than others. For example, when a combination of MSC and standard normal variate (SNV) was used, RMSECV and RMSEP were the lowest at 0.8551 and 0.8561 and $r_c^2$ and $r_p^2$ were the highest at 0.8533 and 0.6546; wavelength ranges of 360-380, 546-690, 760, 915, 931-939, 942, 953, 971, 978, 981, 988, and 992-1019 nm were most influential for SSC determination. The PLSR model with the spectral value of the corresponding region confirmed that the RMSEP decreased to 0.6841 and $r_p^2$ increased to 0.7795 as compared to the values of the entire wavelength band. In this study, we confirmed the feasibility of using a hyperspectral scanner image obtained from outdoors for the SSC measurement of apples. These results indicate that the application of field data and sensors could possibly expand in the future.

• Korean Journal of Remote Sensing
• /
• v.37 no.2
• /
• pp.321-335
• /
• 2021

Particulate matter (PM10 and PM2.5 with a diameter less than 10 and 2.5 ㎛, respectively) can be absorbed by the human body and adversely affect human health. Although most of the PM monitoring are based on ground-based observations, they are limited to point-based measurement sites, which leads to uncertainty in PM estimation for regions without observation sites. It is possible to overcome their spatial limitation by using satellite data. In this study, we developed machine learning-based retrieval algorithm for ground-level PM10 and PM2.5 concentrations using aerosol parameters from Geostationary Ocean Color Imager (GOCI) satellite and various meteorological parameters from a numerical weather prediction model during January to December of 2019. Gradient Boosted Regression Trees (GBRT) and Light Gradient Boosting Machine (LightGBM) were used to estimate PM concentrations. The model performances were examined for two types of feature sets-all input parameters (Feature set 1) and a subset of input parameters without meteorological and land-cover parameters (Feature set 2). Both models showed higher accuracy (about 10 % higher in R2) by using the Feature set 1 than the Feature set 2. The GBRT model using Feature set 1 was chosen as the final model for further analysis(PM10: R2 = 0.82, nRMSE = 34.9 %, PM2.5: R2 = 0.75, nRMSE = 35.6 %). The spatial distribution of the seasonal and annual-averaged PM concentrations was similar with in-situ observations, except for the northeastern part of China with bright surface reflectance. Their spatial distribution and seasonal changes were well matched with in-situ measurements.

• Korean Journal of Soil Science and Fertilizer
• /
• v.43 no.6
• /
• pp.968-974
• /
• 2010

We assessed the feasibility of discrete wavelet transform (DWT) applied for the spectral processing to enhance the estimation performance quality of soil organic matters using visible-near infrared spectra and mapped their distribution via block Kriging model. Continuum-removal and $1^{st}$ derivative transform as well as Haar and Daubechies DWT were used to enhance spectral variation in terms of soil organic matter contents and those spectra were put into the PLSR (Partial Least Squares Regression) model. Estimation results using raw reflectance and transformed spectra showed similar quality with $R^2$ > 0.6 and RPD> 1.5. These values mean the approximation prediction on soil organic matter contents. The poor performance of estimation using DWT spectra might be caused by coarser approximation of DWT which not enough to express spectral variation based on soil organic matter contents. The distribution maps of soil organic matter were drawn via a spatial information model, Kriging. Organic contents of soil samples made Gaussian distribution centered at around 20 g $kg^{-1}$ and the values in the map were distributed with similar patterns. The estimated organic matter contents had similar distribution to the measured values even though some parts of estimated value map showed slightly higher. If the estimation quality is improved more, estimation model and mapping using spectroscopy may be applied in global soil mapping, soil classification, and remote sensing data analysis as a rapid and cost-effective method.

• Korean Journal of Remote Sensing
• /
• v.38 no.6_1
• /
• pp.1181-1189
• /
• 2022

Clouds are composed of tiny water droplets, ice crystals, or mixtures suspended in the atmosphere and cover about two-thirds of the Earth's surface. Cloud detection in satellite images is a very difficult task to separate clouds and non-cloud areas because of similar reflectance characteristics to some other ground objects or the ground surface. In contrast to thick clouds, which have distinct characteristics, thin transparent clouds have weak contrast between clouds and background in satellite images and appear mixed with the ground surface. In order to overcome the limitations of transparent clouds in cloud detection, this study conducted cloud detection focusing on transparent clouds using machine learning techniques (Random Forest [RF], Convolutional Neural Networks [CNN]). As reference data, Cloud Mask and Cirrus Mask were used in MOD35 data provided by MOderate Resolution Imaging Spectroradiometer (MODIS), and the pixel ratio of training data was configured to be about 1:1:1 for clouds, transparent clouds, and clear sky for model training considering transparent cloud pixels. As a result of the qualitative comparison of the study, bothRF and CNN successfully detected various types of clouds, including transparent clouds, and in the case of RF+CNN, which mixed the results of the RF model and the CNN model, the cloud detection was well performed, and was confirmed that the limitations of the model were improved. As a quantitative result of the study, the overall accuracy (OA) value of RF was 92%, CNN showed 94.11%, and RF+CNN showed 94.29% accuracy.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.37 no.6
• /
• pp.563-570
• /
• 2019

Red tide has caused massive fish kills in Korean coastal waters with devastating economic loss in the aquaculture industry since 1995. Remote sensing technique has shown to be effective for the detection of red tide in wide areas, where the absorption property of red tide water plays a central role in understanding the red tide reflectance. This study analyzed the optical absorption property of sea waters dominated by the dinoflagellate specie of Alexandirum affine, off the Tongyeong area in August, 2017. Water samples collected from 20 stations in the ship-based campaign were measured for absorption by pigment, suspended solid, and dissolved organic matter, with the corresponding water quality variables such as chlorophyll concentration and total suspended solid. The analysis showed that Alexandrium-dominated water exhibits strong absorption in the spectral range below 400 nm unlike that of diatom-dominated waters, and greater fluctuations in the range of 400 nm - 500 nm. The packaging effect in pigment absorption was stronger in Alexandrium-dominated waters, and the exponent in the absorption by detritus and gelbstoff is disparate for diatom and Alexandrium. In the model for the detritus and gelbstoff absorption (a_dg(λ)=a_dg(λ₀)e^-s(λ-λ₀)), the optimal exponent coefficient(s) for the Alexandrium was close to 0.01 rather than to 0.015, which was commonly use for modelling diatom waters.

• KOREAN JOURNAL OF CROP SCIENCE
• /
• v.50 no.4
• /
• pp.238-246
• /
• 2005

Rice yield and protein content have been shown to be highly variable across paddy fields. In order to characterize this spatial variability of rice within a field, two-year experiments were conducted in 2002 and 2003 in a large-scale rice field of $6,600m^2$ In year 2004, an experiment was conducted to know if variable rate treatment (VRT) of N fertilizer, that was prescribed for site-specific management at panicle initiation stage, could reduce spatial variation in yield and protein content of rice while increasing yield compared to conventional uniform N topdressing (UN, 33kg N/ha at PIS) method. VRT nitrogen prescription for each grid was calculated based on the nitrogen (N) uptake (from panicle initiation to harvest) required for target rice protein content of $6.8\%$, natural soil N supply, and recovery of top-dressed N fertilizer. The required N uptake for target rice protein content was calculated from the equations to predict rice yield and protein content from plant growth parameters at panicle initiation stage (PIS) and N uptake from PIS to harvest. This model· equations were developed from the data obtained from the previous two-year experiments. The plant growth parameters for the calculation of the required N were predicted non-destructively by canopy reflectance measurement. Soil N supply for each grid was obtained from the experiment of year 2003, and N recovery was assumed to be $60\%$ according to the previous reports. The prescribed VRT N ranged from 0 to 110kg N/ha with an average of 57kg/ha that was higher than 33 kg/ha of UN. The results showed that VRT application successfully worked not only to reduce spatial variability of rice yield and protein content but also to increase rough rice yield by 960kg/ha. The coefficient of variation (CV) for rice yield and protein content was reduced significantly to $8.1\%$ and $7.1\%$ in VRT from $14.6\%$ and $13.0\%$ in UN, respectively. And also the average protein content of milled rice in VRT showed very similar value of target protein content of $6.8\%$. In conclusion the procedure used in this paper was believed to be reliable and promising method for reducing within-field spatial variability of rice yield and protein content. However, inexpensive, reliable, and fast estimation methods of natural N supply and plant growth and nutrition status should be prepared before this method could be practically used for site-specific crop management in large-scale rice field.