• 대한원격탐사학회지
• /
• 제38권5_1호
• /
• pp.511-521
• /
• 2022

위성영상에서의 구름 탐지 및 제거는 지형관측과 분석을 위해 필수적인 과정이다. 임계값 기반의 구름탐지 기법은 구름의 물리적인 특성을 이용하여 탐지하므로 안정적인 성능을 보여주지만, 긴 연산시간과 모든 채널의 영상 및 메타데이터가 필요하다는 단점을 가지고 있다. 최근 활발히 연구되고 있는 딥러닝을 활용한 구름탐지 기법은 4개 이하의 채널(RGB, NIR) 영상만을 활용하고도 짧은 연산시간과 우수한 성능을 보여주고 있다. 본 논문에서는 해상도가 다른 이종 데이터 셋을 활용하여 학습데이터 셋에 따른 딥러닝 네트워크 성능 의존도를 확인하였다. 이를 위해 DeepLabV3+ 네트워크를 구름탐지의 채널 별 특징이 추출되도록 개선하고 공개된 두 이종 데이터 셋과 혼합 데이터로 각각 학습하였다. 실험결과 테스트 영상과 다른 종류의 영상으로만 학습한 네트워크에서는 낮은 Jaccard 지표를 보여주었다. 그러나 테스트 데이터와 동종의 데이터를 일부 추가한 혼합 데이터로 학습한 네트워크는 높은 Jaccard 지표를 나타내었다. 구름은 사물과 달리 형태가 구조화 되어 있지 않아 공간적인 특성보다 채널 별 특성을 학습에 반영하는 것이 구름 탐지에 효과적이므로 위성 센서의 채널 별 특징을 학습하는 것이 필요하기 때문이다. 본 연구를 통해 해상도가 다른 이종 센서의 구름탐지는 학습데이터 셋에 매우 의존적임을 확인하였다.

• 한국광학회:학술대회논문집
• /
• 한국광학회 2005년도 하계학술발표회
• /
• pp.108-109
• /
• 2005

• 한국지리정보학회지
• /
• 제15권4호
• /
• pp.1-14
• /
• 2012

하천은 육지 표면에서 일정한 물길을 따라 흐르는 물줄기를 의미하며, 하천 매핑 작업은 하천유역의 지형 변화 연구 및 하천 유역의 홍수 모니터링 연구 등에 매우 중요한 역할을 한다. 그러나 하천의 수위변화로 인한 유역 내 지표면의 수위 및 유량의 불균일성 등으로 인하여, 기존의 지반조사 기술은 하천 매핑 작업에 효과적이지 않다. 공간 정보 자료는 해당 지역에 접근하지 않고도 해당 지역에 관한 지형적인 정보를 획득할 수 있어서, 하천 지형 조사 및 하천 측량 등 하천 유역의 지형연구에 굉장히 유용하게 쓰일 수 있다. 본 연구에서는, 각각의 다른 파라미터를 사용하여 영상분류 기법 중의 하나인 ISODATA(Iterative Self_Organizing Data Analysis) 분류기법을 적용하여 RapidEye 영상으로부터 하천을 추출하는 방법을 제시하였다. 우선, RapidEye 영상으로부터 NIR(Near InfraRed) 밴드 영상과 NDVI(Normalized Difference Vegetation Index) 영상을 생성한 뒤, 이를 각각의 파라미터로 설정한다. 생성된 각각의 영상에 ISODATA 기법을 적용한 뒤, 후처리 과정을 통하여 각각의 영상으로부터 하천을 추출하도록 한다. 각각의 영상에서 추출한 하천의 경계선 또한 Sobel 에지 추출 기법을 통하여 추출된다. 점검 점들을 이용하여 정확도 검증을 수행한 결과, NIR 밴드로부터 추출한 하천의 정확도가 NDVI 영상으로부터 추출한 하천의 정확도보다 더 높다는 것을 알 수 있다.

• 한국근적외분광분석학회:학술대회논문집
• /
• 한국근적외분광분석학회 2001년도 NIR-2001
• /
• pp.1247-1247
• /
• 2001

Constituents of animal biofluids such as milk, blood and urine contain information specifically related to metabolic and health status of the ruminant animals. Some changes in composition of biofluids can be attributed to disease response of the animals. Mastitis is a major problem for the global dairy industry and causes substantial economic losses from decreasing milk production and reducing milk quality. The purpose of this study was to investigate potential of NIRS combined with multivariate analysis for cow's mastitis diagnosis based on NIR spectra of milk, blood and urine. A total of 112 bulk milk, urine and blood samples from 4 Holstein cows were analyzed. The milk samples were collected from morning milking. The urine samples were collected before morning milking and stored at -35$^{\circ}C$ until spectral analysis. The blood samples were collected before morning milking using a catheter inserted into the carotid vein. Heparin was added to blood samples to prevent coagulation. All milk samples were analyzed for somatic cell count (SCC). The SCC content in milk was used as indicator of mastitis and as quantitative parameter for respective urine and blood samples collected at same time. NIR spectra of blood and milk samples were obtained by InfraAlyzer 500 spectrophotometer, using a transflectance mode. NIR spectra of urine samples were obtained by NIR System 6500 spectrophotometer, using 1 mm sample thickness. All samples were divided into calibration set and test set. Class variable was assigned for each sample as follow: healthy (class 1) and mastitic (class 2), based on milk SCC content. SIMCA was implemented to create models of the respective classes based on NIR spectra of milk, blood or urine. For the calibration set of samples, SIMCA models (model for samples from healthy cows and model for samples from mastitic cows), correctly classified from 97.33 to 98.67% of milk samples, from 97.33 to 98.61% of urine samples and from 96.00 to 94.67% of blood samples. From samples in the test set, the percent of correctly classified samples varied from 70.27 to 89.19, depending mainly on spectral data pretreatment. The best results for all data sets were obtained when first derivative spectral data pretreatment was used. The incorrect classified samples were 5 from milk samples,5 and 4 from urine and blood samples, respectively. The analysis of changes in the loading of first PC factor for group of samples from healthy cows and group of samples from mastitic cows showed, that separation between classes was indirect and based on influence of mastitis on the milk, blood and urine components. Results from the present investigation showed that the changes that occur when a cow gets mastitis influence her milk, urine and blood spectra in a specific way. SIMCA allowed extraction of available spectral information from the milk, urine and blood spectra connected with mastitis. The obtained results could be used for development of a new method for mastitis detection.

• 한국근적외분광분석학회:학술대회논문집
• /
• 한국근적외분광분석학회 2001년도 NIR-2001
• /
• pp.1253-1253
• /
• 2001

The ability to accurately assess wine quality is important during the wine making process, particularly when allocating batches of wines to styles determined by consumer requirements. Grape payments are often determined by the quality category of the wine that is produced from them. Wine quality, in terms of sensory characteristics, is normally a subjective measure, performed by experienced winemakers, wine competition judges or winetasting panellists. By nature, such assessments can be biased by individual preferences and may be subject to day-to-day variation. Taste and aroma compounds are often present in concentrations below the detection limit of near infrared (NIR) spectroscopy but the more abundant organic compounds offer potential for objective quality grading by this technique. Samples were drawn from one of Australia's major wine shows and from BRL Hardy's post-vintage wine quality allocation tastings. The samples were scanned in transmission mode with a FOSS NIR Systems 6500, over the wavelength range 400-2500 ㎚. Data analysis was performed with the Vision chemometrics package. With samples from the allocation tastings, the best correlations between NIR spectra and tasting data were obtained with dry red wines. These calibrations used loadings in the wavelengths related to anthocyanins, ethanol and possibly tannins. Anthocyanins are a group of compounds responsible for colour in red wines - restricting the wavelengths to those relating to anthocyanins produced calibrations of similar accuracy to those using the full wavelength range. This was particularly marked with Merlot, a variety that tends to have relatively lower anthocyanin levels than Cabernet Sauvignon and Shiraz. For dry white wines, calibrations appeared to be more dependent on ethanol characteristics of the spectrum, implying that quality correlated with fruit maturity. The correlations between NIR spectra and sensory data obtained using the wine show samples were less significant in general. This may be related to the fact that within most classes in the show, the samples may span vintages, glowing areas and winemaking styles, even though they may be made from only one grape variety. For dry red wines, the best calibrations were obtained with a class of Pinot Noir - a variety that tends to be produced in limited areas in Australia and would represent the least matrix variation. Good correlations were obtained with a tawny port class - these wines are sweet, fortified wines, that are aged for long periods in wooden barrels. During the ageing process Maillard browning compounds are formed and the water is lost through the barrels in preference to ethanol, producing “concentrated” darkly coloured wines with high alcohol content. These calibrations indicated heaviest loadings in the water regions of the spectrum, suggesting that “concentration” of the wines was important, whilst the visible and alcohol regions of the spectrum also featured as important factors. NIR calibrations based on sensory scores will always be difficult to obtain due to variation between individual winetasters. Nevertheless, these results warrant further investigation and may provide valuable Insight into the main parameters affecting wine quality.

• 지적과 국토정보
• /
• 제48권1호
• /
• pp.71-91
• /
• 2018

최근 영농분야에서 종자파종, 병충해 방제 등에 무인항공기(UAV ; Unmanned Aerial Vehicle)를 활용한 응용이 활발히 진행되고 있다. 본 연구는 UAV에 다양한 파장대의 영상센서를 탑재하고 SfM(Structure from Motion) 영상해석기법과 연계한'고해상 저고도 원격탐측시스템(UAS ; Unmanned Aerial System)'를 구성, UAS 기반 식생조사의 효용성을 고찰하여 정밀영농의 활용성을 검토하였다. 이를 위해 저가 UAV에 가시 컬러(VIS_RGB ; Visible Red, Green, and Blue) 영상센서, 수정된 BG_NIR(Blue Green_Near Infrared Red) 근적외 영상 센서, $7.5{\sim}13.5{\mu}m$ 분광대역의 열적외 영상(TIR ; Thermal Infrared Red)센서를 조합 연계한 UAS를 구성하였다. 또한, 가시 근적외 및 열적외 파장대를 기본요소로 광합성에 따른 식물의 엽록소, 질소 및 수분 함유량 등을 검토할 수 있는 총 10종의 식생지수를 선정, 식생상태 검출에 활용하였다. 시험대상지에 대한 각 파장대역의 영상을 획득하고 사전에 조사된 지상 피복현황을 기준으로 각 식생지수의 분포도 및 식생지수 간 상관성(결정계수 R2) 등을 비교 고찰하여 무인항공기를 활용한 가시 컬러, 근 적외 및 열 적외 영상에 의한 식생상태의 검측 수행능력을 검토하였다. 저가 무인항공기에 VIS_RGB, BG_NIR 및 TIR 영상 센서를 탑재, 식생조사의 효용성을 종합적으로 검토한 결과, 인공위성과 항공영상에 의존한 과거의 식생조사방식 대비, 영상해상도, 경제성 및 운용성 면에서 UAV기반 고해상 저고도 원격탐측시스템(UAS)의 효용성을 입증할 수 있었으므로 정밀농업, 수계 및 산림조사 등의 분야에 그 활용이 기대된다.

• 한국근적외분광분석학회:학술대회논문집
• /
• 한국근적외분광분석학회 2001년도 NIR-2001
• /
• pp.1156-1156
• /
• 2001

This work aimed to prove the feasibility of NIR spectroscopy to detect vegetable protein isolates (soy, pea and wheat) in milk powder. Two hundred and thirty-nine samples of genuine and adulterated milk powder (NIZO, Ede, NL) were analysed by NIRS using an InfraAlyzer 500 (Bran+Luebbe). NIR spectra were collected at room temperature, and data were processed by using Sesame Software (Bran+Luebbe). Separated calibrations for each non-milk protein added, in the range of 0-5%, were calculated. NIR data were processed by using Sesame Software (Bran+Luebbe). Prediction and validation were made by using a set of samples not included into the calibration set. The best calibrations were obtained by the PLSR. The type of data pre-treatment (normalisation, 1$\^$st/ derivative, etc..) was chosen to optimize the calibration parameters. NIRS technique was able to predict with good accuracy the percentage of each vegetable protein added to milk powder (soy: R$^2$ 0.994, SEE 0.193, SEcv 0.301, RMSEPall 0.148; pea: R$^2$ 0.997, SEE 0.1498, SEcv 0.207, RMSEPall 0.148, wheat: R$^2$ 0.997, SEE 0.1418, SEcv 0.335, RMSEPall 0.149). Prediction results were compared to those obtained using other two techniques: capillary electrophoresis and competitive ELISA. On the basis of the known true values of non-vegetable protein contents, the NIRS was able to determine more accurately than the other two techniques the percentage of adulteration in the analysed samples.

• 한국근적외분광분석학회:학술대회논문집
• /
• 한국근적외분광분석학회 2001년도 NIR-2001
• /
• pp.1158-1158
• /
• 2001

Fast and dynamic biochemical, enzymatic and morphological changes occur during the so-called generative development and during the vegetative processes in seeds. The most characteristic biochemical and compositional changes of this period are the formation and decline of storage components or their precursors, the change of their degree in polymerization and an extensive change in water content. The aim of the present study was to detect the maturation processes in seed nondestructively and to verify the applicability of near infrared spectroscopic methods in the measurement of physiological, chemical and biochemical changes in wheat seed. The amount and variation of different water “species” has been changed intensively during maturation. Characteristic changes of three water absorption bands (1920, 1420 and 1150 nm) during maturation were analysed. It was concluded that the free/bound transition of water molecules could be followed sensitively in different region of NIR spectra. Kinetic changes of carbohydrate reserves were characteristic during maturation. An intensive formation and decline of carbohydrate reserves were observed during early stage of maturation (0 -13 days, high energy demand). An accelerated formation of storage carbohydrates (starch) was detected in the second phase of maturation. Five characteristic absorption bands were analysed which were sensitive indicators the changes of carbohydrates occurred during maturation. Precursors of protein synthesis and the synthesis of reserve proteins and their kinetic changes during maturation were followed from NIR spectra qualitative and qualitatively. Dynamic formation of amino acids and the changes of N forms were detected by spectroscopic, chromatographic and by capillary electrophoresis methods. Calibration equations were developed and validated in order to measure the optimal maturation time protein and moisture content of developing wheat seeds. The spectroscopic methods are offering chance and measurement potential in order to detect fine details of physiological processes. The spectra have many hidden details, which can help to understand the biochemical background of processes.

• 한국토양비료학회지
• /
• 제36권2호
• /
• pp.92-103
• /
• 2003

Research and technological advances in the field of remote sensing have greatly enhanced the ability to detect and quantify physical and biological stresses that affect the productivity of agricultural crops. Reflectance in specific visible and near-infrared regions of the electromagnetic spectrum have proved useful in detection of nutrient deficiencies. Especially crop canopy sensors as a ground remote sensing measure the amount of light reflected from nearby surfaces such as leaf tissue or soil and is in contrast to aircraft or satellite platforms that generate photographs or various types of digital images. Multi-spectral vegetation indices derived from crop canopy reflectance in relatively wide wave band can be used to monitor the growth response of plants in relation to environmental factors. The normalized difference vegetation index (NDVI), where NDVI = (NIR-Red)/(NIR+Red), was originally proposed as a means of estimating green biomass. The basis of this relationship is the strong absorption (low reflectance) of red light by chlorophyll and low absorption (high reflectance and transmittance) in the near infrared (NIR) by green leaves. Thereafter many researchers have proposed the other indices for assessing crop vegetation due to confounding soil background effects in the measurement. The green normalized difference vegetation index (GNDVI), where the green band is substituted for the red band in the NDVI equation, was proved to be more useful for assessing canopy variation in green crop biomass related to nitrogen fertility in soils. Consequently ground remote sensing as a non destructive real-time assessment of nitrogen status in plant was thought to be useful tool for site specific crop nitrogen management providing both spatial and temporal information.

• Journal of Photoscience
• /
• 제9권2호
• /
• pp.515-517
• /
• 2002

Photodynamic therapy (PDT) is a cancer treatment modality which utilizes the cytotoxicity of the active singlet oxygen derived from irradiation of a tumor accumulated photosensitizer. As the oxygen in the singlet state radiates an emission of 1270nm wavelength when it decays to the triplet state, detection of the emission helps us to understand the mechanism of PDT or to evaluate photosensitizers. We detected the 1270nm emission from photosensitizers Photofrin and ATX-SI0 in vitro and in vivo by means of high sensitive NIR detectors. We obtained the maximum amount of singlet oxygen at irradiation wavelength of 665-670nm from a HeLa tumor in a nude mouse which is injected with ATX-S10.