請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/78945
標題: | 以無人機獲取之多光譜影像建立田間作物性狀調查暨栽培管理決策支援系統 Developing a Field Crop Phenotyping and Management Decision Support System with Unmanned Aerial Vehicle-derived Multispectral Images |
作者: | Tien-Cheng Wang 王天成 |
指導教授: | 盧虎生(Huu-Sheng Lur) |
共同指導教授: | 蔡育彰(Yu-Chang Tsai) |
關鍵字: | 植生指數,多光譜影像,無人機,田間性狀調查,決策支援系統,雜草, vegetative indices,multispectral images,UAV,field phenotyping,decision support system,weed, |
出版年 : | 2018 |
學位: | 碩士 |
摘要: | 本試驗旨於建立一套標準程序可用於無人機多光譜影像作物性狀時序調查及田間栽培管理決策支援系統。本試驗場域位在桃園區農改場水稻 (Oryza sativa L.) 氮肥試驗田,試驗期作包含2017共兩期作及2018一期作。透過地面調查,觀察不同栽培品種其氮素、產量及穀粒品質性狀對於不同氮肥處理之反應; 同時由無人機攜載多光譜與數位相機時序蒐集空拍影像,測試四個波段、四十三種植生指標、積分指標與性狀之相關性。結果顯示,氮素、產量和穀粒品質等性狀,其最佳配適植生指數會隨不同的品種、期作或植生指數取自的部位而改變。單一品種性狀與植生指數迴歸會比混合品種有更好的表現。與桃園三號在2017年產量最相關植生指標為取自葉部的MCARI/MTVI2 (一期與二期作: R2adjusted為0.462及0.735;移植後第一百二十三天及九十八天)。與桃園三號在2017年之穀粒品質最相關之植生指數,為來自穗部和穗葉混合之red edge 頻段(中心735 nm) 反射光譜 (一期與二期作: R2adjusted為0.887及0.883;移植後第一百二十三天及九十八天)。穀粒品質性狀圖譜(由混合十個品種之穀粒品質分數與移植後第一百二十三天綠光波段(中心550 nm)建立之回歸關係(R2adjusted為0.439)所得),可作為篩選單一品系,甚至是單一植株表現的新工具。另外,首次於本研究提出雜草及穗部專一性指標,可用於分辨作物、雜草以及稻穗。在抽穗期將葉與穗分離可以改善植生指標與調查性狀迴歸分析的決定係數,突顯在成熟期利用穗部專一指標分離穗部對於改善性狀調查精準度的重要性。雜草專一指標可以在栽培早期量化水稻田間雜草覆蓋面積與提供光譜特徵。綜合本研究各項結果顯示,本系統除可從多時序多光譜影像中提供植物專一植生指數之外,亦可提供作物、雜草、穗部覆蓋面積等資訊做為作物栽培管理決策之重要參考。 The aim of this study is to develop a standard protocol for field UAV time series multispectral images phenotyping and a management decision support system. Rice (Oryza sativa L.) nitrogen fertilizer treatment field located in Taoyuan District Agricultural Research and Extension Station (TYDARES). Plant nitrogen status, yield, and grain quality phenotype of different varieties as well as UAV derived multispectral images were collected from in 2017 and 2018. Forty-three vegetative indices (VIs) derived from four original wavebands reflectance and integration of VIs were tested in this study for regression analysis. Result shows that the best VIs for different phenotypes regression varies over time and varieties. Correlation between single variety’s VI and phenotype was higher than pooled phenotypes and VIs. Grain yield of rice variety TY3 were highly correlated with MCARI/MTVI2 at 123 days after transplanting (DAT) (R2adjusted = 0.462) and 98 DAT (R2adjusted = 0.735) from leaf part in the first and second crop season in 2017. Grain quality score of TY3 were highly correlated with the reflectance of red edge (center 735 nm) at 123 DAT (R2adjusted = 0.887) and 98 DAT (R2adjusted = 0.883) from panicle and mixture of panicle and leaf part in the first and second crop season in 2017. Grain quality map, derived from regression of grain quality score of pooled ten varieties with reflectance of green band (center 550) at 123 DAT (R2adjusted = 0.439), offers a new way for single variety or even individual plant performance screening. In this study, we proposed two novel VIs, panicle specific VI (PSVI) and weed specific VI (WSVI). PSVI strongly improves the correlation of VIs with grain quality and leaf nitrogen phenotype at heading stage. Furthermore, WSVI successfully detected and localized weeds in rice field at early vegetative stage of the second crop season in 2017, providing valuable information as a supporting reference for field management decision. In summary, this system can provide not only time-series plant specific vegetative indices profiles, but also cover area and spectral feature profiles specific to crop, weed, and panicle, which is an essential reference information for crop management decision. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/78945 |
DOI: | 10.6342/NTU201803558 |
全文授權: | 有償授權 |
電子全文公開日期: | 2023-08-23 |
顯示於系所單位: | 農藝學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-107-R05621114-1.pdf 目前未授權公開取用 | 8.18 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。