| 摘要: | 對泰國而言，水稻是一個穩定且重要的出口經濟作物，泰國也是最重要水稻出口國之一。隨著人口的增加及氣候變遷，水稻監測在許多國家已經變成一個重要的議題。在泰國，水稻監測是非常重要且必需的，因為水稻是國家最重要的經濟及糧食作物，全國約有百分之六十的農夫的主要種植作物為水稻。對於大面積的水稻種植地區而言，以人力調查進行水稻監測為一耗時且成本高的調查方式，近年來，使用遙測技術進行水稻監測則可以較低的成本、較高的效率取得相當精確地監測成果。
大城府省(Ayutthaya)省位於泰國中部，是泰國重要的水稻生產省分。依據水稻生長物候，泰國中部的水稻生長有兩期(溼季與乾季)，水稻在溼季生長時是多雲且多雨的，不利於使用光學衛星進行監測任務；近年來，合成孔徑雷達衛星在水稻判識亦扮演重要的角色，由於合成孔徑雷達衛星以主動式微波訊號進行地面觀測，微波訊號可以穿透雲層，在多雲覆蓋的天候仍有極佳的監測能力，過去研究亦指出合成孔徑雷達適合對於在雲雨天氣下種植的水稻進行監測。
本研究針對泰國中部大城府省地區，利用2015年(5月-11月)溼季及2016年(5月-11月)乾季的Sentinel-1 VH極化影像，監測並繪製該省的水稻種植範圍。本研究主要分為以下四個步驟：第一步，將原始影像進行前處理，包括輻射校正、幾何校正及雜訊濾除；第二步，配合水稻物候，並考慮研究區內耕地種作、收成時間的差異性，利用多時序合成孔徑雷達影像計算動態NDSI(Normalized Difference Sigma Naught Index)；第三步驟，以expectation–maximization (EM)演算法訂定閾值進行水稻分類；第四步，利用地真資料進行判釋成果的精度評估。研究成果顯示溼季水稻分類結果的總體精度與Kappa係數分別為80.1%與0.60；乾季水稻分類結果的總體精度為83.3%，Kappa係數為0.67。研究結果顯示Sentinel-1 VH極化影像對於泰國中部水稻的判識有相當好的成果，而本研究提出之水稻判釋方法具有相當好的應用潛力。
;Rice is staple food and an economically important export crop of Thailand. Thailand, therefore, is one of the largest rice producers and exporters in the world. With the increase in population and climate change, rice monitoring has become an essential issue around the world. The rice monitoring in Thailand is highly significant and indispensable, as rice is the country’s most important crop, and around 60 percent of Thai farmers mainly plant rice. The rice monitoring is also necessary for economic and ecological development planning. Traditionally, statistics data from the field survey was used to map rice area and estimate rice production. However, the field surveying is time consuming and costly. Thus, remote sensing monitoring has been widely used instead due to its higher effectiveness in rice mapping, using the optical image or synthetic aperture radar (SAR) image. Following the rice crop calendar, there are two rice growing seasons (wet season and dry season) in the Central of Thailand, a very important rice production region in the country. Rice growing season is in the rainy season with cloudy weather which limits the optical sensors usually affected by cloud cover. Hence, the optical sensors are difficult to apply for the rice monitoring in cloudy wet-season. Recently, SAR imagery has played a more important role in rice crop mapping. Many studies indicated that SAR has been a suitable tool for monitoring rice cultivated in cloudy and rainy weather because it uses microwave radiation which can pass through clouds.
The aim of this study is to map the rice-growing areas in Ayutthaya province, Central Thailand using Sentinel-1 VH-polarized SAR images during the wet season (May-November) in 2015 and the dry season (November-April) in 2016. Four essential steps were required to conduct the study: (1) image pre-processing, including radiometric correction, geometric correction, and speckle noise removal, (2) Dynamic Normalized Difference Sigma Naught Index (NDSI) calculation, using time series SAR images to better detect rice fields when they have different growing time, (3) rice growing area classification with applying a threshold derived from expectation–maximization (EM) algorithm, and (4) mapping accuracy assessment, by comparing with ground reference data. The classification result for the wet-season rice indicated the overall accuracy of 80.1% and Kappa coefficient of 0.60. For the dry-season rice, the overall accuracy and kappa coefficient were 83.3% and 0.67, respectively. This study demonstrates the potential of the applicability of Sentinel-1 VH-polarized SAR imagery for rice area mapping in Central Thailand. |
