http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
YOLO를 이용한 SAR 영상의 선박 객체 탐지: 편파별 모델 구성과 정확도 특성 분석
임윤교,윤유정,강종구,김서연,정예민,최소연,서영민,이양원,Yungyo Im,Youjeong Youn,Jonggu Kang,Seoyeon Kim,Yemin Jeong,Soyeon Choi,Youngmin Seo,Yangwon Lee 대한원격탐사학회 2023 大韓遠隔探査學會誌 Vol.39 No.5
Ship detection at sea can be performed in various ways. In particular, satellites can provide wide-area surveillance, and Synthetic Aperture Radar (SAR) imagery can be utilized day and night and in all weather conditions. To propose an efficient ship detection method from SAR images, this study aimed to apply the You Only Look Once Version 5 (YOLOv5) model to Sentinel-1 images and to analyze the difference between individual vs. integrated models and the accuracy characteristics by polarization. YOLOv5s, which has fewer and lighter parameters, and YOLOv5x, which has more parameters but higher accuracy, were used for the performance tests (1) by dividing each polarization into HH, HV, VH, and VV, and (2) by using images from all polarizations. All four experiments showed very similar and high accuracy of 0.977 ≤ AP@0.5 ≤ 0.998. This result suggests that the polarization integration model using lightweight YOLO models can be the most effective in terms of real-time system deployment. 19,582 images were used in this experiment. However, if other SAR images,such as Capella and ICEYE, are included in addition to Sentinel-1 images, a more flexible and accurate model for ship detection can be built.
오픈소스 하드웨어와 딥러닝 기반 객체 탐지 알고리즘을 활용한 교내 유동인구 분석
김보람,임윤교,신실,이진혁,추성원,김나경,박미소,윤홍주 한국전자통신학회 2022 한국전자통신학회 논문지 Vol.17 No.1
In this study, the floating population survey and analysis of Pukyong National University campus were conducted using images through an object detection algorithm based on the open source hardware Raspberry Pi and deep learning technology. For the study, images were collected for a total of 5 days from May 10th to May 14th using Raspberry Pi. After that, people were detected from the collected images using YOLO3's IMAGEAI and YOLOv5 models, and Haar-like features and HOG models were used for accuracy comparison analysis. As a result of comparison, the smallest floating population was observed on the 10th day, the anniversary of the opening of the school, and the largest floating population was observed on the 12th day for the entrance and the 14th for the exit. If the spatial and temporal scope of the study is expanded, it is expected that more accurate floating population analysis will be possible.
Sentinel-1 레이더 식생지수와 AutoML을 이용한Sentinel-2 NDVI 결측화소 복원
윤유정,강종구,김서연,정예민,최소연,임윤교,서영민,원명수,천정화,김경민,장근창,임중빈,이양원 대한원격탐사학회 2023 大韓遠隔探査學會誌 Vol.39 No.6
위성영상 기반의 정규식생지수(normalized difference vegetation index, NDVI)는 넓은 영역에서 주기적인정보를 수집할 수 있어 산림 및 농업 모니터링에 주로 사용된다. 그러나 광학센서 기반 식생지수는 구름 등의영향으로 일부 지역에서 결측을 가지기 때문에, 본 연구는 전천후 및 주야에 관계없이 관측 가능한 Sentinel-1의합성 개구 레이더(synthetic aperture radar, SAR) 영상을 활용하여 Sentinel-2 NDVI 결측값을 복원하는 모델을개발하였다. 이는 광학적으로 관측이 어려운 구름 조건이나 야간에도 NDVI를 추정할 수 있는 잠재력을 보여준다. Automated machine learning (AutoML)을 활용한 비선형 결측복원모델의 5폴드(fold) 교차검증 결과, 절대오차 7.214E-05, 상관계수 0.878의 NDVI 복원 성능을 보였다. 이를 통해 시공간 연속적인 NDVI 생산 방법론을발전시켜, 전천후 식생 모니터링에 필요한 정보 생산에 기여할 수 있을 것으로 기대된다.
딥러닝과 Landsat 8 영상을 이용한 캘리포니아 산불 피해지 탐지
서영민,윤유정,김서연,강종구,정예민,최소연,임윤교,이양원 대한원격탐사학회 2023 大韓遠隔探査學會誌 Vol.39 No.6
The increasing frequency of wildfires due to climate change is causing extreme loss of lifeand property. They cause loss of vegetation and affect ecosystem changes depending on their intensityand occurrence. Ecosystem changes, in turn, affect wildfire occurrence, causing secondary damage. Thus,accurate estimation of the areas affected by wildfires is fundamental. Satellite remote sensing is used forforest fire detection because it can rapidly acquire topographic and meteorological information aboutthe affected area after forest fires. In addition, deep learning algorithms such as convolutional neuralnetworks (CNN) and transformer models show high performance for more accurate monitoring of fireburntregions. To date, the application of deep learning models has been limited, and there is a scarcityof reports providing quantitative performance evaluations for practical field utilization. Hence, this studyemphasizes a comparative analysis, exploring performance enhancements achieved through both modelselection and data design. This study examined deep learning models for detecting wildfire-damagedareas using Landsat 8 satellite images in California. Also, we conducted a comprehensive comparisonand analysis of the detection performance of multiple models, such as U-Net and High-ResolutionNetwork-Object Contextual Representation (HRNet-OCR). Wildfire-related spectral indices such asnormalized difference vegetation index (NDVI) and normalized burn ratio (NBR) were used as inputchannels for the deep learning models to reflect the degree of vegetation cover and surface moisturecontent. As a result, the mean intersection over union (mIoU) was 0.831 for U-Net and 0.848 for HRNet-OCR, showing high segmentation performance. The inclusion of spectral indices alongside the base wavelength bands resulted in increased metric values for all combinations, affirming that the augmentationof input data with spectral indices contributes to the refinement of pixels. This study can be applied toother satellite images to build a recovery strategy for fire-burnt areas.
6SV2.1과 GK2A AOD를 이용한 기계학습 기반의 Sentinel-2 영상 대기보정
김서연,윤유정,강종구,정예민,최소연,임윤교,서영민,박찬원,이경도,나상일,안호용,류재현,이양원,Seoyeon Kim,Youjeong Youn,Jonggu Kang,Yemin Jeong,Soyeon Choi,Yungyo Im,Youngmin Seo,Chan-Won Park,Kyung-Do Lee,Sang-Il Na,Ho-Yong Ahn,Jae-Hyun Ryu 대한원격탐사학회 2023 大韓遠隔探査學會誌 Vol.39 No.5
In this letter, we simulated an atmospheric correction for Sentinel-2 images, of which spectral bands are similar to Compact Advanced Satellite 500-4 (CAS500-4). Using the second simulation of the satellite signal in the solar spectrum - vector (6SV)2.1 radiation transfer model and random forest (RF), a type of machine learning, we developed an RF-based atmospheric correction model to simulate 6SV2.1. As a result, the similarity between the reflectance calculated by 6SV2.1 and the reflectance predicted by the RF model was very high.