Application of UAV images to rice growth status analysis

利用遙測技術於農業需求上，為避免遙測影像空間解析度不足及被雲層遮避之影響，大多使用無人飛行載具 (Unmanned Aerial Vehicle, UAV) 取得高解析度的遙測影像，其優勢為高機動性並能有效取代人眼的觀測結果。本論文使用 UAV 以進行稻田拍攝，並儲存可見光影像、 近紅外光影像及航高資訊，使用常態化差異植生指標 (Normalization Difference Vegetation Index, NDVI) 及水稻覆蓋面積與基地面積的比值 (綠覆率) 來進行水稻生長分析。 因 UAV 空拍影像會受不可控制的環境因素造成影像上形成光斑。本論文利用影像處理方式以排除光斑位置，進而提高綠覆率計算之正確性，其中影像處理程序依序為，在可見光影像上二值化選定光斑範圍，以紅光波段及近紅外光波段作為NDVI影像，光斑範圍於NDVI影像做幾何對位以排除光斑位置。本論文以行政院農業委員會臺中區農業改良場提供研究區域為嘉義縣溪口鄉之水稻田，收集葉面積量測及植株乾重數據與本論文綠覆率進行比較，透過數據分析以評估綠覆率與兩者數據的相關性。 由研究數據分析，以各水稻田區塊之綠覆率進行線性分析，結果得知綠覆率及葉面積於低投入之永續農耕 (Sustainable Agro-ecosystem, SA) 相關係數為 0.82 ，而高投入之慣行農耕 (Conventional Agro-ecosystem, CA)相關係數為 0.95 ，則綠覆率與植株乾重於 SA相關係數為0.92、CA相關係數為0.90。水稻結穗 6 月影像綠覆率趨勢下降之原因推測為葉片枯萎，實地調查後葉面積及植株乾重為明顯下滑與推測相符。本論文以UAV影像有效地證明水稻生長分析情況。Precision agriculture uses remote sensing images to predict crop yield. To avoid the lack of spatial resolution of remote sensing images or the influence of cloud occlusion, an unmanned aerial vehicle (UAV) is commonly used to obtain high-resolution remote sensing images. With its high mobility, a UAV can acquire effectively images to replace the observations of human. This thesis uses UAVs to acquire visible light images and near-infrared images of rice at specific flight altitudes. Normalization difference vegetation index (NDVI) and the ratio of rice coverage to base area (green coverage rate) are employed for the analysis of rice growth. With uncontrollable environmental factors, UAV aerial images are contaminated with light spots sometimes. This thesis proposes an image processing method, including the binarization of the selected spot area on the visible light images and the alignment of the NDVI images to eliminate the area of the light spots, to improve the correctness of green coverage rate. The research area is the rice field in Xikou Township, Chiayi County governed by the Taichung District Agricultural Research and Extension Station, Council of Agriculture, Executive Yuan. The collected leaf area measurement and plant dry weight data are compared with the green coverage of this paper to assess the correlation between green coverage and the two data. The research data suggest that the reason for the decline in the green coverage rate in June is the leaves being withered. The leaf area measurement and plant dry weight of rice decreased significantly in image-derived data and the on-site investigation. The green coverage rate was positively correlated with leaf area and plant dry weight. The correlation coefficient between the green coverage rate and leaf area is 0.82 and 0.95 for sustainable agro-ecosystem (SA) and the conventional agro-ecosystem (CA), respectively, whereas the correlation coefficient between the green coverage rate and plant dry weight is 0.92 and 0.90 for SA and CA, respectively. This thesis proves UAV images effectively in analysis of rice growth.誌謝 i 摘要 ii ABSTRACT iii 目錄 iv 表目錄 vi 圖目錄 vii 第一章 緒論 1 1.1 研究動機 1 1.2 研究目的 2 1.3 論文架構 2 第二章 文獻回顧 3 2.1 水稻生育之影響 3 2.2 二值化 4 2.3 常態化差異植生指標(Normalization Difference Vegetation Index, NDVI) 5 2.4 綠覆率 6 2.5 葉面積 6 第三章 研究方法與理論 8 3.1 二值化 9 3.2 常態化差異植生指標 (Normalization Difference Vegetation Index, NDVI) 10 3.3 綠覆率(Ratio of green cover) 11 3.4 葉面積 11 第四章 成果與討論 12 4.1 研究材料 12 4.2 研究區域概述 13 4.3 常態化差異植生指標(NDVI) 18 4.4 二值化 23 4.5 綠覆率與葉面積及植株乾重真值 26 4.6 成果與分析 36 4.7 小結 41 第五章 結論與建議 42 5.1 結論 42 5.2 建議 43 參考文獻 4