| 摘要: | 水稻為台灣重要的糧食作物之一，然而，因為農業技術進步、天然災害頻仍、飲食習慣改變等因素，耕作面積與產量預估成為一個重要的議題，為了不讓稻米供需失衡，有效的監測稻米收穫面積成為一項重要的任務，過往以人力探訪的方式來監測稻米顯得費時費力，現今，利用免費的衛星資料可以達成有效率的大範圍面積監測，因此本研究區域選擇位於臺灣中部，傳統農業密集的雲林縣，也是主要生產稻米的縣市之一。
歐洲太空總署哥白尼計劃的C波段合成孔徑雷達衛星Sentinel-1自2014年起提供免費影像，並且其1A以及1B 兩顆衛星在台灣地區的再訪週期為6天拍攝一次，較頻繁的時間解析度對於偵測作物的短時序變化有很大的助益。本研究利用Sentinel-1於2019年拍攝60張雷達影像的兩種極化(VH與VV)後向散射係數，以其時間序列辨識水稻的物候特徵，並且利用小波分析將VH、VV以及VV/VH比值平滑化，分析水稻特定的物候特徵。本研究以像素為單位，分別找出水稻種植開始季節(SOS)、分櫱末期(EOT)、以及收穫季節(EOS)的時間點，藉此定義水稻的生長週期，最後比對水稻田的現地樣本與每一像素的時間序列之相關性。此外，也透過添加地籍資料作為物件式分類的幾何特徵來提高準確度。
結果顯示雲林縣整體Kappa值為0.64，從單一鄉鎮來分析各鄉鎮之間水稻的種植特性與分布特性，Kappa值較高鄉鎮落在斗六市、斗南鎮、大埤鄉等，由於其地區種植密度最高;而最低值落在土庫鎮、元長鄉等，由於其境內多數地區有混合種植不同作物的情形，以及耕作面積過小的因素影響。此外，從雲林整體Kappa值與地籍面積的相關係數來看，面積1000平方公尺是本方法的最小判釋單位。;Paddy rice is one of the leading agricultural products in Taiwan, especially in Yunlin County. However, ground-based monitoring activities are highly time- and resource-consuming, so a few have been deployed on a large scale. Among various spaceborne sensors, radar satellite data provide timely and wide-area information without affecting by the cloud. Hence, this study aims to retrieve time series of paddy-specific phenology using Sentinel-1 Synthetic Aperture Radar (SAR) data and to map parcels for paddy rice in Yunlin. Our study processed 60 SAR images in 2019, and follow three major steps: (1) Extracting specific paddy phenology curves in training sites, by the temporal behavior of SAR backscattering coefficients (VH, VH, and the ratio of VV/VH) and using Wavelet transform to remove the noise; (2) Defining the start of the season (SOS), end of tillering (EOT), and end of the season (EOS) in paddy growing cycle through backscattered signal curve; (3) Comparing the cross-correlation coefficients with the specific paddy phenology curves. Besides, we add the cadastral map as the geometric feature for object-based classification to improve accuracy. In our validation, the kappa value of Yunlin is 0.64. The overall accuracy is over 0.7 for all townships in Yunlin. The region with the high kappa value is in the eastern part of Yunlin (Douliu City, Dounan, and Dapi Township, etc.) because of the high density of the rice field. On the other hand, the region with the low kappa value is in the western part of Yunlin (Tuku and Yuanchang Township, etc.) due to crop diversification. Furthermore, the whole Yunlin’s Pearson correlation coefficient (r) compared with the kappa index and the size of the area indicates the area of 1000 square meters is the smallest unit of detecting rice fields in our study. |
