• Korean Journal of Remote Sensing
• /
• v.35 no.6_1
• /
• pp.959-971
• /
• 2019

CNN (Convolutional Neural Network) is one representative deep learning algorithm, which can extract high-level spatial and spectral features, and has been applied for hyperspectral image classification. However, one significant drawback behind the application of CNNs in hyperspectral images is the high dimensionality of the data, which increases the training time and processing complexity. To address this problem, several CNN based hyperspectral image classification studies have exploited PCA (Principal Component Analysis) for dimensionality reduction. One limitation to this is that the spectral information of the original image can be lost through PCA. Although it is clear that the use of PCA affects the accuracy and the CNN training time, the impact of PCA for CNN based hyperspectral image classification has been understudied. The purpose of this study is to analyze the quantitative effect of PCA in CNN for hyperspectral image classification. The hyperspectral images were first transformed through PCA and applied into the CNN model by varying the size of the reduced dimensionality. In addition, 2D-CNN and 3D-CNN frameworks were applied to analyze the sensitivity of the PCA with respect to the convolution kernel in the model. Experimental results were evaluated based on classification accuracy, learning time, variance ratio, and training process. The size of the reduced dimensionality was the most efficient when the explained variance ratio recorded 99.7%~99.8%. Since the 3D kernel had higher classification accuracy in the original-CNN than the PCA-CNN in comparison to the 2D-CNN, the results revealed that the dimensionality reduction was relatively less effective in 3D kernel.

• Korean Journal of Remote Sensing
• /
• v.28 no.4
• /
• pp.347-355
• /
• 2012

A fast algorithm for endmember extraction is proposed in this study which extracts min. and max. pixels from each band after MNF transform as candidate pixels for endmember. This method finds endmembers not from the entire image pixels but only from the previously extracted candidate pixels. The experimental results by N-FINDR using a simulated hyperspectral image data and AVIRIS Cuprite image data showed that the proposed fast algorithm extracts the same endmembers with the conventional methods. More studies on the effect of noise and more adaptive criteria in extracting candidate pixels are expected to increase the usability of this method for more fast and efficient analysis of hyperspectral image data.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2021.06a
• /
• pp.89-89
• /
• 2021

부영양화된 호수나 유속이 느린 하천에서 발생하는 녹조의 과도한 발생은 하천 생태계 훼손, 동식물의 건강, 담수의 오염 등 환경 사회 경제적으로 큰 피해를 준다. 현재 수질 측정망은 정해진 지점에서 Chlorophyll-a(Chl-a), Phycocyanin(PC)을 대표농도로 산정하고 조류경보에 활용하고 있으나, 일주일에 한번씩 샘플링을 통해 Chl-a 및 PC를 측정하여 시공간적인 신뢰성의 문제가 제기될 수 있다. 본 연구에서는 기존 점단위 조류 모니터링의 한계점을 개선하기 위해 초분광영상 자료를 머신러닝 기법에 적용하여 Chl-a 및 PC 산정 알고리즘을 개발하였다. 이를 위해 Chl-a와 PC의 최대 흡수, 반사 파장대, 주요 물 흡수 파장대 자료를 조합하여 9개의 파장비를 구축하였으며, 기존 연구에서 활용한 머신러닝 기법인 Partial Least Square, Random Forest, Gradient Boosting, Support Vector Machine, K-Nearest Neighbor, Artificial Neural Network를 검토하여 최적 모델을 선정하였다. 학습된 머신러닝의 성능을 R², NSE, RMSE 목적함수를 이용해 평가하였으며, 그 결과 ANN이 각각 PC 0.801, 0.755, 11.774 mg/m³, Chl-a 0.733, 0.622, 8.736 mg/m³로 가장 우수한 성능을 보였다. 최적화 된 ANN 모델을 백제보 상류 2016-2017년 항공 초분광영상에 적용하여 시공간에 따른 조류 분포변화를 평가하고자 한다.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2022.05a
• /
• pp.89-89
• /
• 2022

하천 합류부에 지천이 유입되는 경우 복잡한 3차원적 흐름 구조를 발생시키고 이로 인해 유사혼합 및 지형 변화가 활발히 발생하게 된다. 특히, 하천 합류부에서 부유사 거동은 하천의 세굴과퇴적, 하천 지형 변화, 하천 생태계, 하천구조물 안정성 등에 직접적으로 영향을 미치기 때문에 이에 대한 정확한 분석이 하천 관리 및 재해 예방에 필수적인 요소이다. 기존의 하천 합류부 부유사 계측 자료들은 재래식 채취 방식으로 수행되어 시공간적 해상도가 매우 낮아서 실측 자료만으로 합류부에서 부유사 혼합을 분석하기에는 한계가 존재하기에 대하천의 부유사 혼합 거동 해석에 수치모형이 주로 활용되어 왔다. 본 연구에서는 하천 합류부에서 부유사 거동을 공간적으로 정밀하게 분석하기 위해 드론 기반초분광 영상을 활용하여 하천 합류부에 최적화된 부유사 계측 방법론을 제시하였다. 현장에서 계측한 초분광 자료와 부유사 농도간의 관계를 구축하기 위하여 기계학습모형인 랜덤포레스트(Random Forest) 회귀 모형과 합류부에서 분광 특성이 다른 두 하천의 특성을 정확하게 반영하기 위한 가우시안 혼합 모형 (Gaussian Mixture Model) 기반 초분광 군집화 기법을 결합하였다. 본 연구에서 구축한 방법론을 낙동강과 황강의 합류부에 적용한 결과, 초분광 군집을 통해 두하천 흐름의 경계층을 명확히 구별하였으며, 이를 바탕으로 지류와 본류에 대해 각각 분리된 회귀 모형을 구축하여 복잡한 합류부 근역 경계층에서의 부유사 거동을 보다 정확하게 재현하였다. 또한 나아가서 재현된 고해상도의 부유사 공간분포를 바탕으로 경계층에서 강한 두 흐름이 혼합되어 발생한 와류(Wake)가 부유사 혼합에 미치는 영향을 규명하였고, 하천 합류부에서 발생하는 전단층의 수평방향 대규모 와류가 부유사 혼합 양상에 지배적 영향을 미치는 것으로 확인하였다.

• Journal of the Korean Society of Marine Environment & Safety
• /
• v.28 no.7
• /
• pp.1120-1128
• /
• 2022

In the event of a marine accident, the longer the exposure time to the sea increases, the faster the chance of survival decreases. However, because the search area of the sea is extremely wide compared to that of land, marine object detection technology based on the sensor mounted on a satellite or an aircraft must be applied rather than ship for an efficient search. The purpose of this study was to rapidly detect an object in the ocean using a hyperspectral image sensor mounted on an aircraft. The image captured by this sensor has a spatial resolution of 8,241 × 1,024, and is a large-capacity data comprising 127 spectra and a resolution of 0.7 m per pixel. In this study, a marine object detection model was developed that combines a seawater identification algorithm using DBSCAN and a density-based land removal algorithm to rapidly analyze large data. When the developed detection model was applied to the hyperspectral image, the performance of analyzing a sea area of about 5 km² within 100 s was confirmed. In addition, to evaluate the detection accuracy of the developed model, hyperspectral images of the Mokpo, Gunsan, and Yeosu regions were taken using an aircraft. As a result, ships in the experimental image could be detected with an accuracy of 90 %. The technology developed in this study is expected to be utilized as important information to support the search and rescue activities of small ships and human life.

• Journal of Korean Society for Geospatial Information Science
• /
• v.25 no.1
• /
• pp.9-17
• /
• 2017

Hyperspectral imagery is used in the land cover classification with the principle component analysis and minimum noise fraction to reduce the data dimensionality and noise. Recently, studies on the supervised classification using various features having spectral information and spatial characteristic have been carried out. In this study, principle component bands and normalized difference vegetation index(NDVI) was utilized in the supervised classification for the land cover classification. To utilize additional information not included in the principle component bands by the hyperspectral imagery, we tried to increase the classification accuracy by using the NDVI. In addition, the extended attribute profiles(EAP) generated using the morphological filter was used as the input data. The random forest algorithm, which is one of the representative supervised classification, was used. The classification accuracy according to the application of various features based on EAP was compared. Two areas was selected in the experiments, and the quantitative evaluation was performed by using reference data. The classification accuracy of the proposed algorithm showed the highest classification accuracy of 85.72% and 91.14% compared with existing algorithms. Further research will need to develop a supervised classification algorithm and additional input datasets to improve the accuracy of land cover classification using hyperspectral imagery.

• Korean Journal of Remote Sensing
• /
• v.30 no.1
• /
• pp.61-73
• /
• 2014

This study reviewed an application of water quality estimation using an Airborne Hyperspectral Imagery (A-HSI) and tested a part of Han River water quality (especially suspended solid) estimation with available in-situ data. The estimation of water quality was processed two methods. One is using observation data as downwelling radiance to water surface and as scattering and reflectance into water body. Other is linear regression analysis with water quality in-situ measurement and upwelling data as at-sensor radiance (or reflectance). Both methods drive meaningful results of RS estimation. However it has more effects on the auxiliary dataset as water quality in-situ measurement and water body scattering measurement. The test processed a part of Han River located Paldang-dam downstream. We applied linear regression analysis with AISA eagle hyperspectral sensor data and water quality measurement in-situ data. The result of linear regression for a meaningful band combination shows $-24.847+0.013L_{560}$ as 560 nm in radiance (L) with 0.985 R-square. To comparison with Multispectral Imagery (MSI) case, we make simulated Landsat TM by spectral resampling. The regression using MSI shows -55.932 + 33.881 (TM1/TM3) as radiance with 0.968 R-square. Suspended Solid (SS) concentration was about 3.75 mg/l at in-situ data and estimated SS concentration by A-HIS was about 3.65 mg/l, and about 5.85mg/l with MSI with same location. It shows overestimation trends case of estimating using MSI. In order to upgrade value for practical use and to estimate more precisely, it needs that minimizing sun glint effect into whole image, constructing elaborate flight plan considering solar altitude angle, and making good pre-processing and calibration system. We found some limitations and restrictions such as precise atmospheric correction, sample count of water quality measurement, retrieve spectral bands into A-HSI, adequate linear regression model selection, and quantitative calibration/validation method through the literature review and test adopted general methods.

• Korean Journal of Remote Sensing
• /
• v.31 no.5
• /
• pp.361-370
• /
• 2015

Various algorithms such as Chronochrome(CC), Principle Component Analysis(PCA), and spectral unmixing have been studied for hyperspectral change detection. Change detection by spectral unmixing offers useful information on the nature of the change compared to the other change detection methods which provide only the locations of changes in the scene. However, hyperspectral change detection by spectral unmixing is still in an early stage. This research proposed a new approach to extract endmembers, which have identical properties in temporally different images, by Iterative Error Analysis (IEA) and Spectral Angle Mapper(SAM). The change map obtained from the difference of abundance efficiently showed the changed pixels. Simulated images generated from Compact Airborne Spectrographic Imager (CASI) and Hyperion were used for change detection, and the experimental results showed that the proposed method performed better than CC, PCA, and spectral unmixing using N-FINDR. The proposed method has the advantage of automatically extracting endmembers without prior information, and it could be applicable for the real images composed of many materials.

• Journal of the Korean Association of Geographic Information Studies
• /
• v.21 no.1
• /
• pp.35-45
• /
• 2018

Image data collected by an airborne hyperspectral camera system have a great usability in coastal line mapping, detection of facilities composed of specific materials, detailed land use analysis, change monitoring and so forh in a complex coastal area because the system provides almost complete spectral and spatial information for each image pixel of tens to hundreds of spectral bands. A few approaches after classifying by a few approaches based on SAM(Spectral Angle Mapper) supervised classification were applied for extracting optimal land cover information from hyperspectral images acquired by CASI-1500 airborne hyperspectral camera on the object of a coastal area which includes both land and sea water areas. We applied three different approaches, that is to say firstly the classification approach of combined land and sea areas, secondly the reclassification approach after decompostion of land and sea areas from classification result of combined land and sea areas, and thirdly the land area-only classification approach using atmospheric correction images and compared classification results and accuracies. Land cover classification was conducted respectively by selecting not only four band images with the same wavelength range as IKONOS, QuickBird, KOMPSAT and GeoEye satelllite images but also eight band images with the same wavelength range as WorldView-2 from 48 band hyperspectral images and then compared with the classification result conducted with all of 48 band images. As a result, the reclassification approach after decompostion of land and sea areas from classification result of combined land and sea areas is more effective than classification approach of combined land and sea areas. It is showed the bigger the number of bands, the higher accuracy and reliability in the reclassification approach referred above. The results of higher spectral resolution showed asphalt or concrete roads was able to be classified more accurately.

• Korean Journal of Remote Sensing
• /
• v.24 no.2
• /
• pp.125-131
• /
• 2008

Hyperspectral imaging is used in a wide variety of research since the image is obtained with a wider wavelength range and more bands than multispectral imaging. However, there are limitations, namely that each band has a shorter wavelength range, the computation cost is increased in the case of numerous bands, and a high correlation between each band and noise bands exists. The previous analysis method does not produce ideal results due to these limitations. Therefore, in the case of using the hyperspectral image, image analysis after eliminating noise bands is more accurate and efficient. In this study, noise band elimination of the hyperspectral image preprocessing is highlighted, and we use fractal dimension for noise band elimination. The Triangular Prism Method is used, being the typical fractal dimension method of the curved surface. The fractal dimension of each band is calculated. We then apply the Continuum Removal method to normalize. A total of 35 bands are estimated by noise band with a threshold value that is obtained empirically. The hyperion hyperstpectral image collected on the EO-1 satellite is used in this study. The result delineates that noise bands of the hyperion image are able to be eliminated with the fractal dimension and Continuum Removal method.