• 대한원격탐사학회지
• /
• 제36권5_3호
• /
• pp.1095-1107
• /
• 2020

홍수 발생 시 위성영상을 활용하여 침수된 지역을 추출하는 것은 홍수 발생 기간 내의 위성영상 취득과 영상에 나타난 침수구역의 정확한 분류 등에서 많은 어려움이 존재한다. 딥러닝은 전통적인 영상분류기법들에 비해 보다 정확도가 높은 위성영상분류기법으로 주목받고 있지만, 광학영상에 비해 홍수 발생 시 위성영상의 취득이 용이한 SAR 영상의 분류 잠재력은 아직 명확히 규명되지 않았다. 본 연구는 대표적인 의미론적 영상 분할을 위한 딥러닝 모델인 SegNet과 U-Net을 활용하여 동남아시아의 라오스, 태국, 필리핀의 대표적인 홍수 발생지역인 코랏 유역(Khorat basin), 메콩강 유역(Mekong river basin), 카가얀강 유역(Cagayan river basin)에 대해 Sentinel-1 A/B 위성영상으로부터 침수지역 추출을 실시하였다. 분석결과 침수지역 탐지에서 SegNet의 Global Accuracy, Mean IoU, Mean BF Score는 각각 0.9847, 0.6016, 0.6467로 나타났으며, U-Net의 Global Accuracy, Mean IoU, Mean BF Score는 각각 0.9937, 0.7022, 0.7125로 나타났다. 국지적 분류결과 확인을 위한 육안검증에서 U-Net이 SegNet에 비해 보다 높은 분류 정확도를 보여주었지만, 모델의 훈련에 필요한 시간은 67분 17초와 187분 19초가 각각 소요되어 SegNet이 U-Net에 비해 약 3배 정도 빠른 처리속도를 보여주었다. 본 연구의 결과는 향후 딥러닝 기법을 활용한 SAR 영상기반의 홍수탐지 모델과 실무적으로 활용이 가능한 자동화된 딥러닝 기반의 수계탐지 기법의 제시를 위한 중요한 참고자료로 활용될 수 있을 것으로 판단된다.

• 한국측량학회지
• /
• 제35권5호
• /
• pp.423-430
• /
• 2017

Recently, data analysis research has been carried out using the deep learning technique in various fields such as image interpretation and/or classification. Various types of algorithms are being developed for many applications. In this paper, we propose a precipitation prediction algorithm based on deep learning with high accuracy in order to take care of the possible severe damage caused by climate change. Since the geographical and seasonal characteristics of Korea are clearly distinct, the meteorological factors have repetitive patterns in a time series. Since the LSTM (Long Short-Term Memory) is a powerful algorithm for consecutive data, it was used to predict precipitation in this study. For the numerical test, we calculated the PWV (Precipitable Water Vapor) based on the tropospheric delay of the GNSS (Global Navigation Satellite System) signals, and then applied the deep learning technique to the precipitation prediction. The GNSS data was processed by scientific software with the troposphere model of Saastamoinen and the Niell mapping function. The RMSE (Root Mean Squared Error) of the precipitation prediction based on LSTM performs better than that of ANN (Artificial Neural Network). By adding GNSS-based PWV as a feature, the over-fitting that is a latent problem of deep learning was prevented considerably as discussed in this study.

• 환경영향평가
• /
• 제8권3호
• /
• pp.77-90
• /
• 1999

The degree of green naturality (DGN) has played a key role for maintaining the environmental quality from inappropriate developments, although the quality and effectiveness of the mapping of DGN has been under debate. In this study, spatial distribution of degree of green naturality was initially estimated from forest stand maps that were produced from the aerial photo interpretation and extensive field survey. Once the boundary of initial classes of DGN were defined, it were overlaid with normalized difference vegetation index (NDVI) data that were derived from the recently obtained Landsat Thematic Mapper data. NDVI was calculated for each pixel from the radiometrically corrected satellite image. There were no significant differences in mean values of vegetation index among the initial DGN classes. However, the satellite derived vegetation index was very effective to delineate the developed and damaged forest lands and to adjust the initial value of DGN according to the distribution of NDVI within each class.

• 대한원격탐사학회지
• /
• 제16권4호
• /
• pp.327-338
• /
• 2000

The forest conditions of North Korea has been a great concern since it was known to be closely related to many environmental problems of the disastrous flooding, soil erosion, and food shortage. To assess the long-term changes of forest area as well as the canopy conditions, several sources of multitemporal satellite data were applied to the study area near Kaesung. KOMPSAT-1 EOC data were overlaid with 1981 topographic map showing the boundaries of forest to assess the deforestation area. Delineation of the cleared forest was performed by both visual interpretation and unsupervised classification. For analyzing the change of forest canopy condition, multiple scenes of Landsat and SPOT data were selected. After preprocessing of the multitemporal satellite data, such as image registration and normalization, the normalized difference vegetation index (NDVI) was derived as a representation of forest canopy conditions. Although the panchromatic EOC data had radiometric limitation to classify diverse cover types, they can be effectively used t detect and delineate the deforested area. The results showed that a large portion of forest land has been cleared for the urban and agricultural uses during the last twenty years. It was also found that the canopy condition of remaining forests has not been improved for the last twenty years. It was also found that the canopy condition of remaining forests has not been improved for the last twenty years. Possible causes of the deforestation and the temporal pattern of canopy conditions are discussed.

• 한국지역지리학회지
• /
• 제2권2호
• /
• pp.183-196
• /
• 1996

지도의 형태는 전통적으로 항공사진이나 위성사진을 이용한 수작업에 의해 추출되어 왔으나, 최근의 원격탐사 기술의 발달은 지도의 형태추출에 새로운 혁신을 가져다 주었다. 공간필터기법을 이용하면 지도형태 추출의 자동화가 가능한데, 특히 선의 추출을 고빈도필터를 이용하여 가능하다. 본 연구에서는 먼저 필터와 지도화의 관계에 대해 이론적으로 고찰한 다음 지도의 선형화와 관련하여 유용하다고 알려진 필터들을 소개하였다. 그리고 이 필터들을 진주시의 SPOT Panchromatic 영상에 적용하였다. 적용한 결과 본고에서 소개한 필터적용 영상이 전반적으로 초기영상보다 개선되어 선형화작업에 매우 유용함을 확인하였다. 특히 초기영상이 포함된 필터영상들의 해석이 용이하였고 선형화가 명확하였다. 그러나 문헌에서 유용성이 인정된 필터라도 일부 필터는 진주시의 경우 선형화작업이 전혀 불가능하였다.

• 대한원격탐사학회:학술대회논문집
• /
• 대한원격탐사학회 2004년도 Proceedings of ISRS 2004
• /
• pp.626-629
• /
• 2004

Texture imaging, which means texture image creation by co-occurrence relation, has been known as one of useful image analysis methodologies. For this purpose, most commercial remote sensing software provides texture analysis function named GLCM (Grey Level Co-occurrence Matrix). In this study, texture-imaging program for GLCM algorithm is newly implemented in the MS Visual IDE environment. While, additional texture imaging modules based on GLDV (Grey Level Difference Vector) are contained in this program. As for GLCM/GLDV texture variables, it composed of six types of second order texture function in the several quantization levels of 2(binary image), 8, and 16: Homogeneity, Dissimilarity, Energy, Entropy, Angular Second Moment, and Contrast. As for co-occurrence directionality, four directions are provided as $E-W(0^{\circ}),\;N-E(45^{\circ}),\;S-W(135^{\circ}),\;and\;N-S(90^{\circ}),$ and W-E direction is also considered in the negative direction of E- W direction. While, two direction modes are provided in this program: Omni-mode and Circular mode. Omni-mode is to compute all direction to avoid directionality problem, and circular direction is to compute texture variables by circular direction surrounding target pixel. At the second phase of this study, some examples with artificial image and actual satellite imagery are carried out to demonstrate effectiveness of texture imaging or to help texture image interpretation. As the reference, most previous studies related to texture image analysis have been used for the classification purpose, but this study aims at the creation and general uses of texture image for urban remote sensing.

• 한국지하수토양환경학회:학술대회논문집
• /
• 한국지하수토양환경학회 2000년도 추계학술대회
• /
• pp.103-106
• /
• 2000

Geophysical data sets from the Chojeong area in the Chungbuk-Do are compositely studied in terms of multi-attribute interpretation for the subsurface mapping of shallow fracture zones, associated with groundwater reservoir. Utilizing a GIS software, the attribute data are implemented to a database; a lineament from the satellite image, electrical resistivities and its standard deviation, radioactivity, seismic velocity, bedrock depth from exploration data. In an attempt to interpret 1-D electrical sounding data in 2-D and 3-D views, 2-D resistivities structures are firstly made by interpolating 1-D plots. Reconstruction of a resistivity volume is found to be an effective scheme for subsurface mapping of shallow fracture zones. Shallow fracture zones in the southeastern part of the study area are commonly correlated in the various exploration data.

• 자원환경지질
• /
• 제33권4호
• /
• pp.321-332
• /
• 2000

The purpose of this study is to analyze landslide that occurred in Yongin area in 1991 using spatial database. For this, landslide locations are detected from aerial photographs interpretation and field survey. The locations of landslide, topography, soil, forest and geology were constructed to spatial database using Geographic Information System (GIS). To establish occurrence factors of landslide, slope, aspect and curvature of topography were calculated from the topographic database. Texture, material, drainage and effective thickness of soil were extracted from the soil database, and type, age, diameter and density of wood were extracted from the forest database. Lithology was extracted from the geological database, and land use was classified from the TM satellite image. Landslide was analyzed using spatial correlation between the landslide and the landslide occurrence factors by bivariate probability methods. GIS was used to analyze vast data efficiently and statistical programs were used to maintain specialty and accuracy. The result can be used to prevention of hazard, land use planning and construction planning as basic data.

• 지구물리와물리탐사
• /
• 제6권1호
• /
• pp.7-12
• /
• 2003

위성영상에 나타나는 선구조는 화소의 명암차에 바탕을 두어 육안판독이나 자동 추출 알고리즘을 이용하여 추출한다. 보통 육안판독은 다양한 보조자료와 연구자의 경험적 지식을 필요로 하기 때문에 더욱 객관적이고 신속한 자동선구조 추출 알고리즘들이 필요하다. 산악지형 외에도 충적층 요소까지 고려하여 개발된 DSTA(dynamic segment tracing algorithm) 알고리즘은 충적층 지역이 산악지형에 대해 비교적 넓은 부분을 차지하고 있는 경우에 충적층 지역에서 노이즈가 나타난다. 본 연구에서는 이러한 노이즈를 감소시킬 수 있는 알고리즘인 AERA(alluvial effect reducing algorithm)를 개발하여 산악지형과 농경지와 도심지역이 넓게 분포되어 있는 지역에 이를 적용하였으며, 기존의 알고리즘만 이용하여 선구조를 추출한 결과와 비교하여 이의 적용 가능성을 알아보았다.

• 대한공간정보학회지
• /
• 제18권2호
• /
• pp.87-92
• /
• 2010

인구의 증가는 하천 유역지역의 토지이용변화를 가속시켜왔으며, 토지이용변화에 대한 공간분포정보는 이들 지역에 대한 효과적인 관리와 계획에 중요한 자료가 되고 있다. 본 연구의 목적은 1989년, 1999년, 2002년도 IRS LISS III 영상자료를 이용하여 인도 서부에 위치한 Moyar유역지역의 1:50,000축척 토지이용변화도를 생성하는데 있다. 약 9가지의 토지이용분류자료는 3개년간의 영상자료를 시각적 판독방법에 의해 추출하였으며, 토지이용변화 검색은 관측시기 I(1989-1999)과 관측시기 II (1999-2002)에 대한 행렬분석방법에 의해 수행되었다. 본 연구를 통하여 기간 II가 기간 I보다 지속가능한 개발과 난개발을 방지하기 위한 실질적인 정보를 보여 주었다.