Publications

Detailed Information

Inter-comparison of image fusion products against in-situ spectral measurements over a heterogeneous rice paddy landscape : 지표면이 이질적인 농경지를 대상으로 영상합성자료와 현장스펙트럼측정 간의 상호비교

Cited 0 time in Web of Science Cited 0 time in Scopus
Authors

공주원

Advisor
류영렬
Major
농업생명과학대학 생태조경·지역시스템공학부(생태조경학)
Issue Date
2019-02
Publisher
서울대학교 대학원
Description
학위논문 (석사)-- 서울대학교 대학원 : 농업생명과학대학 생태조경·지역시스템공학부(생태조경학), 2019. 2. 류영렬.
Abstract
다양한 규모의 생태계 역동성을 관찰하기 위해 중요한 시공간 해상도가 높은 영상에 대한 기술의 발전과 함께 그 정확도 평가가 요구된다. 기존의 영상합성 자료를 평가하는 연구들은 지표면이 이질적인 경관에서 체계적인 평가를 하기 알맞은 픽셀단위의 현장 스펙트럼 측정값이 결여되었다. 본 연구는 식물의 생장기 동안 지표면이 이질적인 벼논 경관에서 현존하는 네가지 영상합성 자료- ESTARFM와 FSDAF, STAIR, CESTEM -를 현장 스펙트럼 측정값과 상호비교하였다. 영상합성의 NDV (정규식생지수)가 시공간적으로 변화하는 것에 대해 실험하였다. 영상합성의 NDVI는 현장 NDVI에 대해 알맞거나 높은 선형성을 보였고, 그 값들이 음의 방향으로 편향되었다 (0.73< R2 <0.93, -8 %
The advances from high spatiotemporal resolution images that are important for the monitoring of ecosystem dynamics across different scales demand the accuracy assessment. Previous evaluation studies of fusion products lack pixel-level in-situ spectral measurements that fit the systematic evaluation in a heterogeneous landscape. We conducted an inter-comparison of existing four image fusion products with systematically collected in-situ spectral measurements over heterogeneous rice paddy landscape during the whole growing season. The four image fusion products include enhanced spatial-temporal adaptive reflectance fusion model (ESTARFM), flexible spatiotemporal data fusion (FSDAF), satellite data integration (STAIR), and CubeSat enabled spatiotemporal enhancement method (CESTEM). We tested fusion NDVI products, Normalized difference vegetation index, in terms of spatial patterns and temporal variations. Fusion NDVI products showed the moderate to strong linear relationships and negative bias against ground NDVI (R2 range 0.73 to 0.93, bias up to -11%). Although fusion NDVI products captured the NDVI variation, ground NDVI had a larger amplitude than fusion NDVI products. The results indicated that the positive bias of input against in-situ measurement caused the overestimation on mixed land cover type, and the negative bias of input against in-situ measurement caused the underestimation on homogeneous rice paddy cover type. Furthermore, fusion NDVI products could underestimate the variations in vegetation activity. The temporal dependency of ESTARFM, FSDAF, and STAIR was not clearly detected on the mixed land cover type. By comparing fusion NDVI products to ground NDVI, the strength of each fusion product was confirmed. We expect our in-situ spectral measurement could be useful in quantifying uncertainty in image fusion products, improving fine resolution cropland mapping and monitoring, and choosing the algorithm depending on the research purpose.
Language
eng
URI
https://hdl.handle.net/10371/150999
Files in This Item:
Appears in Collections:

Altmetrics

Item View & Download Count

  • mendeley

Items in S-Space are protected by copyright, with all rights reserved, unless otherwise indicated.

Share