2015-01-012024-05-17https://scholars.lib.ntu.edu.tw/handle/123456789/684183摘要：於前期計畫中，藉由插秧期前、插秧期、分蘗盛期及收割前等四個不同時段的衛星影像資料進行研究區內休耕作物判釋，於休耕作物之判釋正確性已可達到87~94%，證明在運用多時段地表植生覆蓋時序及光譜特徵變化進行休耕作物監測上，已可接近業務執行之95%正確性需求。本年度計畫以既有之多時段衛星影像資料進行休耕作物判釋技術發展為基礎，除延續以四個不同時段的衛星影像資料，進行休耕作物分布判釋並製作各類型主題圖外，更加入：1.以先進遙測技術(如人飛行載具機動監測)進行水稻及休耕作物之分布範圍判釋並產製各類主題圖；2.評估以無人飛行載具技術進行農作物分布監測之可行性。藉由本年度計畫之研究成果，除可逐步強化休耕作物的遙測判釋關鍵與技術，提昇大範圍休耕作物判釋正確性外，同時並提出以無人飛行載具進行農作物生長狀態機動監測之應用發展方案。在智慧型農田坵塊單元萃取模組部分，以像元式及區塊式資料轉換特徵向量型態分析，以及持續探討不同分類器對休耕田使用型態的判釋能力，持續進行提高坵塊精準性之研究。 <br> Abstract: In the previous year project, with the spectrum change of 4 pair image data from rice transplantation to harvesting, we found the interpretation accuracy of fallow fields is from 87 to 94%, close to the requirement of formal operation of Agriculture and Food Agency (AFA). In this year, according to the previous result, we will still use 4 pair of image data to interpret fallow fields. The improvements in this year are: 1. Applying further remote sensing techniques, such as Unman Aerial Vehicle (UAV) images to interpret rice crops and fallow fields in research area. 2. Evaluate the possibility and usages of applying UAV techniques to monitoring agriculture behaviors in the formal operation of AFA. Base on the results of this project, we can improve the interpreting keys and the accuracy of fallow fields in large area. Moreover, we can also develop the suit methodology of flexible monitoring of agriculture growth status. In smart farmland unit detection model part, we will apply pixelbased and regional-based data transfer feature vector analysis to improve the accuracy of farmland unit detection. After that, this sudy still to explore fifferent classifiers (Decision Tree, Support Vector machine and ANN) for improve the accuracy of farmland unit detection on fallow field’s problems.影像分割影像分類影像單元化知識庫管理休耕田遙測技術無人飛行載具Image SegmentationImage ClassificationImage UnitKnowlage Databased ManagementFallow FieldsRemote Sensing TechnologyUnmanned Aerial Vehicl