TY - JOUR
T1 - Robust and adaptive band-to-band image transform of UAS miniature multi-lens multispectral camera
AU - Jhan, Jyun Ping
AU - Rau, Jiann Yeou
AU - Haala, Norbert
N1 - Funding Information:
This research was financially supported by the Ministry of Science and Technology (MOST), Taiwan (R.O.C.) with project number NSC 103-2119-M-006-002 and MOST 104-2119-M-006-005 . The authors are respectively grateful to Prof. Cho-Ying Huang of Department of Geography, National Taiwan University, Mr. William Lee of Aeroland UAV Inc., and Mr. Kircheis Liu of GEOSAT Aerospace Co., Ltd. for providing a MiniMCA-12, UAS platform, and system integration for collecting dataset (1). The authors are also grateful to Carbon-Based Technology Inc. for providing the UAV platforms and assisting in collecting datasets (3)–(5) and Pix4D Inc. for sharing dataset (6) via the Internet.
Funding Information:
This research was financially supported by the Ministry of Science and Technology (MOST), Taiwan (R.O.C.) with project number NSC 103-2119-M-006-002 and MOST 104-2119-M-006-005. The authors are respectively grateful to Prof. Cho-Ying Huang of Department of Geography, National Taiwan University, Mr. William Lee of Aeroland UAV Inc., and Mr. Kircheis Liu of GEOSAT Aerospace Co., Ltd. for providing a MiniMCA-12, UAS platform, and system integration for collecting dataset (1). The authors are also grateful to Carbon-Based Technology Inc. for providing the UAV platforms and assisting in collecting datasets (3)–(5) and Pix4D Inc. for sharing dataset (6) via the Internet.
Publisher Copyright:
© 2017 International Society for Photogrammetry and Remote Sensing, Inc. (ISPRS)
PY - 2018/3
Y1 - 2018/3
N2 - Utilizing miniature multispectral (MS) or hyperspectral (HS) cameras by mounting them on an Unmanned Aerial System (UAS) has the benefits of convenience and flexibility to collect remote sensing imagery for precision agriculture, vegetation monitoring, and environment investigation applications. Most miniature MS cameras adopt a multi-lens structure to record discrete MS bands of visible and invisible information. The differences in lens distortion, mounting positions, and viewing angles among lenses mean that the acquired original MS images have significant band misregistration errors. We have developed a Robust and Adaptive Band-to-Band Image Transform (RABBIT) method for dealing with the band co-registration of various types of miniature multi-lens multispectral cameras (Mini-MSCs) to obtain band co-registered MS imagery for remote sensing applications. The RABBIT utilizes modified projective transformation (MPT) to transfer the multiple image geometry of a multi-lens imaging system to one sensor geometry, and combines this with a robust and adaptive correction (RAC) procedure to correct several systematic errors and to obtain sub-pixel accuracy. This study applies three state-of-the-art Mini-MSCs to evaluate the RABBIT method's performance, specifically the Tetracam Miniature Multiple Camera Array (MiniMCA), Micasense RedEdge, and Parrot Sequoia. Six MS datasets acquired at different target distances and dates, and locations are also applied to prove its reliability and applicability. Results prove that RABBIT is feasible for different types of Mini-MSCs with accurate, robust, and rapid image processing efficiency.
AB - Utilizing miniature multispectral (MS) or hyperspectral (HS) cameras by mounting them on an Unmanned Aerial System (UAS) has the benefits of convenience and flexibility to collect remote sensing imagery for precision agriculture, vegetation monitoring, and environment investigation applications. Most miniature MS cameras adopt a multi-lens structure to record discrete MS bands of visible and invisible information. The differences in lens distortion, mounting positions, and viewing angles among lenses mean that the acquired original MS images have significant band misregistration errors. We have developed a Robust and Adaptive Band-to-Band Image Transform (RABBIT) method for dealing with the band co-registration of various types of miniature multi-lens multispectral cameras (Mini-MSCs) to obtain band co-registered MS imagery for remote sensing applications. The RABBIT utilizes modified projective transformation (MPT) to transfer the multiple image geometry of a multi-lens imaging system to one sensor geometry, and combines this with a robust and adaptive correction (RAC) procedure to correct several systematic errors and to obtain sub-pixel accuracy. This study applies three state-of-the-art Mini-MSCs to evaluate the RABBIT method's performance, specifically the Tetracam Miniature Multiple Camera Array (MiniMCA), Micasense RedEdge, and Parrot Sequoia. Six MS datasets acquired at different target distances and dates, and locations are also applied to prove its reliability and applicability. Results prove that RABBIT is feasible for different types of Mini-MSCs with accurate, robust, and rapid image processing efficiency.
UR - http://www.scopus.com/inward/record.url?scp=85041420400&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85041420400&partnerID=8YFLogxK
U2 - 10.1016/j.isprsjprs.2017.12.009
DO - 10.1016/j.isprsjprs.2017.12.009
M3 - Article
AN - SCOPUS:85041420400
SN - 0924-2716
VL - 137
SP - 47
EP - 60
JO - ISPRS Journal of Photogrammetry and Remote Sensing
JF - ISPRS Journal of Photogrammetry and Remote Sensing
ER -