• 한국물환경학회지
• /
• 제30권3호
• /
• pp.329-339
• /
• 2014

This study aims to estimate chlorophyll-a concentration in rivers using multi-spectral RapidEye imagery and Spectral Mixture Analysis (SMA) and assess the applicability of SMA for multi-temporal imagery analysis. Comparison between images (acquired on Oct. and Nov., 2013) predicted and ground reference chlorophyll-a concentration showed significant performance statistically with determination coefficients of 0.49 and 0.51, respectively. Two band (Red-RE) model for the October and November 2013 RapidEye images showed low performance with coefficient of determinations ($R^2$) of 0.26 and 0.16, respectively. Also Three band (Red-RE-NIR) model showed different performance with $R^2$ of 0.016 and 0.304, respectively. SMA derived Chlorophyll-a concentrations showed relatively higher accuracy than band ratio models based values. SMA was the most appropriate method to calculate Chlorophyll-a concentration using images which were acquired on period of low Chlorophyll-a concentrations. The results of SMA for multi-temporal imagery showed low performance because of the spatio-temporal variation of each end members. This approach provides the potential of providing a cost effective method of monitoring river water quality and management using multi-spectral imagery. In addition, the calculated Chlorophyll-a concentrations using multi-spectral RapidEye imagery can be applied to water quality modeling, enhancing the predicting accuracy.

• 대한원격탐사학회:학술대회논문집
• /
• 대한원격탐사학회 2007년도 Proceedings of ISRS 2007
• /
• pp.504-507
• /
• 2007

The KOMPSAT-2 satellite is a push-broom system with MSC (Multi Spectral Camera) which contains a panchromatic band and four multi-spectral bands covering the spectral range from 450nm to 900nm. The PAN band is composed of six CCD array with 2528 pixels. And the MS band has one CCD array with 3792 pixels. Raw imagery generated from a push-broom sensor contains vertical streaks caused by variability in detector response, variability in lens falloff, pixel area, output amplifiers and especially electrical gain and offset. Relative radiometric calibration is necessary to account for the detector-to-detector non-uniformity in this raw imagery. Non-uniformity correction (NUC) is that the process of performing on-board relative correction of gain and offset for each pixel to improve data compressibility and to reduce banding and streaking from aggregation or re-sampling in the imagery. A relative gain and offset are calculated for each detector using scenes from uniform target area such as a large desert, forest, sea. In the NUC of KOMPSAT-2, The NUC table for each pixel are divided as HF NUC (high frequency NUC) and LF NUC (low frequency NUC) to apply to few restricted facts in the operating system ofKOMPSAT-2. This work presents the algorithm and process of NUC table generation and shows the imagery to compare with and without calibration.

• Advances in nano research
• /
• 제16권6호
• /
• pp.531-548
• /
• 2024

The study focuses on using remote sensing to gather data about the Earth's surface, particularly in urban environments, using satellites and aircraft-mounted sensors. It aims to develop a classification framework for road targets using multi-spectral imagery. By integrating Convolutional Neural Networks (CNNs) with XGBoost, the study seeks to enhance the accuracy and efficiency of road target identification, aiding urban infrastructure management and transportation planning. A novel aspect of the research is the incorporation of quantum sensors, which improve the resolution and sensitivity of the data. The model achieved high predictive accuracy with an MSE of 0.025, R-squared of 0.85, RMSE of 0.158, and MAE of 0.12. The CNN model showed excellent performance in road detection with 92% accuracy, 88% precision, 90% recall, and an f1-score of 89%. These results demonstrate the model's robustness and applicability in real-world urban planning scenarios, further enhanced by data augmentation and early stopping techniques.

• Journal of Forest and Environmental Science
• /
• 제40권1호
• /
• pp.15-23
• /
• 2024

This research aimed to assess the possibility of detecting forest degradation using time-series satellite imagery and three different deep learning-based change detection techniques. The dataset used for the deep learning models was composed of two sets, one based on surface reflectance (SR) spectral information from satellite imagery, combined with Texture Information (GLCM; Gray-Level Co-occurrence Matrix) and terrain information. The deep learning models employed for land cover change detection included image differencing using the Unet semantic segmentation model, multi-encoder Unet model, and multi-encoder Unet++ model. The study found that there was no significant difference in accuracy between the deep learning models for forest degradation detection. Both training and validation accuracies were approx-imately 89% and 92%, respectively. Among the three deep learning models, the multi-encoder Unet model showed the most efficient analysis time and comparable accuracy. Moreover, models that incorporated both texture and gradient information in addition to spectral information were found to have a higher classification accuracy compared to models that used only spectral information. Overall, the accuracy of forest degradation extraction was outstanding, achieving 98%.

• 대한원격탐사학회:학술대회논문집
• /
• 대한원격탐사학회 2003년도 Proceedings of ACRS 2003 ISRS
• /
• pp.175-177
• /
• 2003

To monitor riparian wetlands as one of complex natural ecosystems using remotely sensed data, we need to concurrently consider vegetation, soil and water which constitute complicated wetland ecosystems. To identify riparian distribution we adopted linear Spectral Mixture Analysis in order to improve identification accuracy of riparian areas. This study has indicated that linear SMA adopting tasseled cap endmember selection is an enhanced routine for Identification of riparian wetlands and phenologically autumn imagery is more appropriate to detect riparian vegetation in the Paldang water catchment area.

• 항공우주기술
• /
• 제10권1호
• /
• pp.149-155
• /
• 2011

세계적 기후온난화와 이상기온현상으로 최근 급변하는 농업환경에 대응하기 위해서는 농작물 작황관리 및 예측시스템의 과학화를 통한 정부차원의 대처능력 개선이 시급하다. 농업분야에서 위성정보의 활용은 고해상도 광학 및 레이더 영상의 상용화와 더불어 정밀농업이라는 새로운 가능성을 열어주고 있다. 본 연구에서는 최근 농업분야에서 주목을 받고 있는 RapidEye 위성영상을 사용하여 우리나라 농촌지역의 토지이용유형별 분광식생지수의 시계열 특성을 살펴보았다. 식생과 비식생지역 간에 뚜렷한 시계열 변화양상이 나타났으며, 식생지역 내에서도 산림 수종별, 논 그룹별로 식생지수의 시계열 변화에 차이가 관찰되었다.

• 대한원격탐사학회지
• /
• 제23권5호
• /
• pp.355-363
• /
• 2007

The 3D DWT(The Three Dimensional Discrete Wavelet Transform) scheme is potentially regarded as useful one on analyzing both spatial and spectral information. Nevertheless, few researchers have attempted to process or classified remotely sensed images using the 3D DWT. This study aims to apply the 3D DWT to the land cover classification of optical and SAR(Synthetic Aperture Radar) images. Then, their results are evaluated quantitatively and compared with the results of traditional classification technique. As the experimental results, the 3D DWT shows superior classification results to conventional techniques, especially dealing with the high-resolution imagery and SAR imagery. It is thought that the 3D DWT scheme can be extended to multi-temporal or multi-sensor image classification.

• 대한원격탐사학회지
• /
• 제19권5호
• /
• pp.381-392
• /
• 2003

Multi-temporal approaches using sequential data acquired over multiple years are essential for satisfactory discrimination between many land-cover classes whose signatures exhibit seasonal trends. At any particular time, the response of several classes may be indistinguishable. A harmonic model that can represent seasonal variability is characterized by four components: mean level, frequency, phase and amplitude. The trigonometric components of the harmonic function inherently contain temporal information about changes in land-cover characteristics. Using the estimates which are obtained from sequential images through spectral analysis, seasonal periodicity can be incorporates into multi-temporal classification. The Normalized Difference Vegetation Index (NDVI) was computed for one week composites of the Advanced Very High Resolution Radiometer (AVHRR) imagery over the Korean peninsula for 1996 ～ 2000 using a dynamic technique. Land-cover types were then classified both with the estimated harmonic components using an unsupervised classification approach based on a hierarchical clustering algorithm. The results of the classification using the harmonic components show that the new approach is potentially very effective for identifying land-cover types by the analysis of its multi-temporal behavior.

• International journal of advanced smart convergence
• /
• 제13권3호
• /
• pp.335-344
• /
• 2024

This study explores advanced machine learning techniques for improving crop yield prediction in smart farming, utilizing multi-temporal spectral data from drone-based multispectral imagery. Conducted in garlic orchards in Andong, Gyeongbuk Province, South Korea, the research examines the effectiveness of various vegetation indices and cutting-edge models, including LSTM, CNN, Random Forest, and XGBoost. By integrating these models with the Analytic Hierarchy Process (AHP), the study systematically evaluates the factors that influence prediction accuracy. The integrated approach significantly outperforms single models, offering a more comprehensive and adaptable framework for yield prediction. This research contributes to precision agriculture by providing a robust, AI-driven methodology that enhances the sustainability and efficiency of farming practices.

• 한국컴퓨터그래픽스학회논문지
• /
• 제25권3호
• /
• pp.21-29
• /
• 2019

본 논문에서는 구조-텍스처 분할 기법을 기반으로 위성영상을 분할 융합하여 공간 해상도를 개선시키는 프레임워크를 제시한다. 위성영상은 센서가 감지하는 파장에 따라 다양한 공간해상도를 가진다. 전정 영상 (panchromatic image)은 일반적으로 높은 공간해상도를 가지지만 단일 흑백컬러를 가지고 있는 반면, 다중분광 영상 (multi-spectral image)나 적외선 영상은 전정 영상에 비해 낮은 공간해상도를 가지지만 다양한 분광 밴드정보와 열 정보를 가지고 있다. 본 논문에서는 다중분광 영상이나 적외선 영상의 공간 해상도를 향상시키기 위해 영상의 디테일이 텍스처 영상에만 존재한다는 것에 착안하여 본 프레임워크를 고안하였다. 고안된 프레임워크에서는 저해상도 영상과 고해상도 영상이 구조 영상과 텍스처 영상으로 분할된 뒤, 저해상도 구조영상은 고해상도 구조 영상을 참조하여 가이디드 필터링 된다. 구조-텍스처 영상 모델에 따라 필터링된 저해상도 영상의 구조 영역과 고해상도 영상의 텍스처 영역을 픽셀 단위로 더해져서 최종 영상이 생성된다. 생성된 영상은 저해상도 영상의 밴드와 고해상도 영상의 디테일을 포함한다. 제시하는 방법은 분광해상도와 공간해상도를 모두 보존할 수 있음을 실험적으로 확인하였다.