請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/73620
標題: | 基於精緻農業空拍影像全景拼接與稻田植株識別 Panoramic stitching based on drone images and plant identification in refinement agriculture |
作者: | Yi-Syuan Ye 葉怡萱 |
指導教授: | 黃乾綱 |
關鍵字: | 影像縫合,影像拼接,空拍機全景拼接,稻田影像處理,秧苗辨識, image stitching,aerial camera,panorama stitching,rice field image processing,plant identification., |
出版年 : | 2019 |
學位: | 碩士 |
摘要: | 民以食為天,無論時代如何進步,我們的生活都跟食物脫離不了關係。從古至今農業都是不可或缺的一個重要產業,以前的時代,耕種一直是十分仰賴人力 的一項工作,以往透過大量人力去播種、耕種、施肥,甚至是消滅蟲害、噴灑農藥。隨著科技進步,面對傳統產業開始出現取代大量人力的技術,也因為亞洲地形的特色,亞洲農地狹小破碎,亞洲地區的栽種方式與歐洲、美洲產生很大的區 別,亞洲地區的農地逐漸變稠密,農地面積破碎,農業轉型朝向精緻化,反觀歐 洲、美洲地廣,農田主要為寬廣無邊際。
隨著科技的進步,大量仰賴人力的農業各種耕種技術,已經可以利用機器代替,過去的研究多半專注於高度很高的衛星空照圖,或著自然街景的全景圖,探討影像如何拼接(Image Stitching),如何透過空拍的熱感應圖(NDVI),近年開始出 現許多使用空拍機拍攝空照圖,分析各個階段耕種的品質、如何監控不同耕種階 段的穩定,卻沒有針對低空飛行機拍攝空照圖的分析與探討。本研究目標是透過 空拍機去拍攝稻田,利用空拍機拍攝的影像去縫合拼接出稻田的全景,並且以連 續的稻田影像,進行特徵點比對,將影像中的秧苗進行標記。本研究的目標是透過拼接疊合與標記,達到自動化標記秧苗的效果,以利降低農業仰賴人力的問題, 改善大量人力去清點觀察稻田耕種的品質好壞,透過空拍機的固定頻率拍攝稻田,將稻田圖縫合拼接的結果,進行不同的觀察與討論,也能判斷稻田生長變化。 以往的研究都是高度極高的衛星圖,進行影像拼接(Image Mosaicing),另一種是距離較近的平面街景影像,本研究針對空拍機的拍攝稻田的影像,觀察稻田影像資料具有衛星圖空照圖與平面街景影像的不同特色,本研究基於上述問題為出發點,研究針對此稻田影像擁有的複合狀況進行縫合拼接,為區別過去研究與本研究的差異,將研究分為兩個部份討論,第一部份討論稻田影像的拼接與現有方法的差異,第二部份的研究將影像分割進行影像秧苗的標記與辨識。 本研究用空拍機影像去拼接連續稻田的位置,並且以連續的稻田影像,進行 特徵點比對,將影像中的秧苗進行標記。本研究的目標是透過拼接疊合與標記,達到自動化標記秧苗的效果。 No matter how the times progress, our lives are not related to food. From the past to the present, agriculture is an indispensable and important industry. In the past, farming has always been a work that relies heavily on manpower. In the past, a large number of people have been planting, cultivating, fertilizing, and even eliminating pests and spraying pesticides. Because of the characteristics of Asian terrain, Asian agricultural land is small and fragmented, and the cultivation methods in Asia are very different from those in Europe and America. The agricultural land in Asia is becoming denser, the agricultural land is broken, and the agricultural transformation is becoming more refined. In contrast, Europe and the Americas Wide, farmland is mainly broad and infinite. Advances in technology can already be replaced by machines. In the past, most of the research focused on satellite aerial photos, or panoramic views of natural streetscapes, to explore how images are stitched (Image Stitching), and how to use aerial snapshots (NDVI). In recent years, many aerial photographs have been taken using aerial cameras to analyze the quality of farming at various stages and how to monitor the stability of different farming stages. However, there is no analysis and discussion of aerial photography for low-altitude aerial cameras. The research goal is to shoot rice fields through aerial photography, use the images taken by the aerial camera to stitch and stitch out the panoramic view of the rice fields, and use the continuous rice field image to perform feature point comparison to mark the seedlings in the image. The goal of the research is to achieve the effect of automating the labeling of seedlings by splicing the overlay and marking, so as to reduce the problem of agriculture depending on the people, improve the quality of the rice cultivation by a large number of people, and shoot the rice fields at a fixed frequency of the aerial camera. The results of stitching and stitching in the rice field are subject to different observations and discussions, and the growth of rice fields can also be judged. In this study, the position of the continuous paddy field was spliced by the image of the aerial camera, and the feature points were compared with the continuous rice field image to mark the seedlings in the image. The goal of this study was to achieve automated labeling of seedlings by splicing the overlays and markers. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/73620 |
DOI: | 10.6342/NTU201903880 |
全文授權: | 有償授權 |
顯示於系所單位: | 工程科學及海洋工程學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-108-1.pdf 目前未授權公開取用 | 6.79 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。