Please use this identifier to cite or link to this item:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/91727
Title: | 無人機影像技術結合作物生長曲線應用於小麥生物量預測 Application of UAV Image Technology Combined with Crop Growth Curves for Wheat Biomass Prediction |
Authors: | 李赫珍 He-Chen Lee |
Advisor: | 劉力瑜 Li-Yu Daisy Liu |
Keyword: | 小麥生物量,無人機影像,顏色指標,生長曲線, wheat biomass,UAV images,color indices,growth curve, |
Publication Year : | 2024 |
Degree: | 碩士 |
Abstract: | 生物量能反映作物生長狀態,也是預測最終產量的中間參數,以低成本、省時的方法預測生物量除了能幫助實現有效率的田間管理,也能因應資金相對不足的研究環境,讓技術更廣泛應用於不同的資源條件和降低使用門檻。本研究以低成本的無人機RGB影像,提取可見光波段的顏色指標預測德國小麥生物量,除此之外結合經驗Logistic生長曲線方法,以生長曲線對齊生物量採樣和影像拍攝時間來對生物量進行調整,並探討顏色指標預測小麥生物量複迴歸模型的表現,最後以生長曲線預測小麥生物量為基礎,提出四個年度調整的應用情境。結果顯示,生物量經生長曲線調整後,模型的決定係數提升,其中HSI顏色指標的模型在調整前、後皆有最好的表現 (R^2=0.6141、0.7310),然而次年驗證的決定係數僅0.28左右。情境二以RGB顏色指標預測生長曲線參數r並重新配適生長曲線,所得的 R^2=0.7092,是最符合非破壞性、節省成本的做法。本研究使用無人機影像技術結合生長曲線的方法在未來仍有許多研究的空間來提升準確度,關於四個情境下不同生育時期相對誤差的結果,未來的種植者可以此作為參考,根據其目的與成本考量選擇最適合的調整方法和種植品種。 Biomass serves as a key indicator reflecting crop growth status and is an intermediate parameter for predicting final yield. Predicting biomass in a cost-effective and time-efficient manner not only facilitates efficient field management but also addresses research environments with limited funding, enabling technology to be widely applied across different resource conditions and lowering entry barriers. In this study, low-cost unmanned aerial vehicle (UAV) RGB images were used to extract color indices in the visible light spectrum for predicting biomass of German wheat. Additionally, the empirical Logistic growth curve method was employed, aligning the time of biomass sampling and image capture to adjust biomass. The study explored the performance of color index predictions for wheat biomass using a multiple linear regression model after applying growth curve adjustment. Finally, building upon the growth curve predictions of wheat biomass, four annual adjustment scenarios are proposed. Results showed that after adjusting with the growth curve, the model's accuracy improved. The model using HSI color indices exhibited the best performance both before and after adjustment (R^2=0.6141 and 0.7310), although the validation accuracy for the subsequent year was around 0.28. Scenario two, predicting growth curve parameter r using RGB color indices and refitting a curve, yielded an R^2 with 0.7092, representing a non-destructive, cost-effective approach. This study utilized drone image technology combined with the growth curve method, suggesting numerous opportunities for future research to enhance accuracy. The relative errors during different growth stages in the four scenarios can serve as a reference for future planters to choose the most suitable adjustment methods and crop varieties based on their goals and cost considerations. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/91727 |
DOI: | 10.6342/NTU202400424 |
Fulltext Rights: | 未授權 |
Appears in Collections: | 農藝學系 |
Files in This Item:
File | Size | Format | |
---|---|---|---|
ntu-112-1.pdf Restricted Access | 3.51 MB | Adobe PDF |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.