請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/83185
標題: | 影像分析與信標輔助定位應用於植生牆的維護 Applications of Image Analysis and Beacon Aided Positioning in Maintenance of Living Wall |
其他標題: | Applications of Image Analysis and Beacon Aided Positioning in Maintenance of Living Wall |
作者: | 尤子澔 Tzu-Hao Yu |
指導教授: | 黃振康 |
關鍵字: | 太陽能,IoT,小型氣象站,信標,指紋特徵定位, solar energy,IoT,small weather station,beacon,fingerprint positioning, |
出版年 : | 2022 |
學位: | 碩士 |
摘要: | 植生牆 (Living Wall)是將植物栽種於牆面,或讓植物依附於牆面或間接結構之上,是各種綠化工程中最經濟快速、效益也最高的選擇。但是高空綠化的環境嚴苛,維護來說是一大工程。為了減少人力維護的成本與風險,本研究旨在展現植生牆自動維護系統的概念,希望未來能讓植生牆的維護走向自動化。自動維護系統包含能源監控、植物影像與環境數據的取得、植物綠葉面積與輪廓的影像處理以及維護裝置的移動與定位。在能源部分,維護裝置藉由太陽能作為電力供給來源,使用微控制器製作具有物聯網 (Internet of Things, IoT) 功能之太陽能電流電壓監控裝置,太陽能板之電流與電壓數據將即時上傳至雲端、同時記錄在記憶卡中,實現太陽能發電的監控與評估。另外,製作小型氣象站於戶外收集溫度、濕度、光照等數據,即時上傳至雲端,完成小區域的環境數據收集。在嵌入式系統有限的運算資源下,利用簡易的影像處理方法完成植物綠葉面積的計算與輪廓的取得,能應用於植物生長狀況的評估,與作為判斷修剪需求的依據。在維護裝置的移動上,製作能自動循跡移動的定位車,模擬裝置維護時的移動,並且能為定位實驗節省約 30% 的時間。使用信標 (Beacon) 進行定位,完成多邊定位法與指紋特徵定位法的實驗,分別得到0.796 m與0.186 m的平均誤差,最後使用神經網絡 (Neural Networks) 對指紋特徵進行訓練,得到0.054 m的平均誤差與80.4%的定位準確率。本研究在植生牆自動維護系統的各領域中皆有實踐與討論,希望能提供一些實驗數據與經驗,並激發更多想法與討論。 Living Walls are aimed to plant plants on the wall or let plants attach to the wall or other indirect structure. It is the most economical, fastest and efficient choice for various green projects. However, the maintenance work of high-altitude greening is harsh. To reduce the cost and risk of maintenance, this study focuses to demonstrate the automatic maintenance system of living wall for the possible automation in the future. The automatic maintenance system included energy monitoring, acquisition of plant images and environmental data, image processing of green leaf area and edge of plant, and movement and positioning of maintenance devices. In terms of energy, the maintenance device used solar energy as a source of power supply. Utilizing a microcontroller to establish a solar current and voltage monitoring device with Internet of Things (IoT) functions, the current and voltage data of the solar panel were uploaded to the cloud in real-time and recorded on the memory card to realize the monitoring and evaluation of solar energy. In addition, a small weather station was built to collect data such as temperature, humidity, and outdoor illuminance. The data was uploaded to the cloud in real-time to complete the collection of environmental data in small regions. Based on the limited computing resources from the embedded system, the simple image processing method was used to complete the calculation of green leaf area and the acquisition of the edge of the plant, which could be applied to the evaluation of the plant growth status and the judgement for pruning needs. For the movement of the maintenance device, a positioning vehicle that could automatically track and move was established to simulate the movement of the device during maintenance, saving about 30% of the time for the positioning experiments. The experiments of the multilateral positioning and fingerprint positioning method was completed, with average errors of 0.796 m and 0.186 m respectively. Finally, the neural network was used to train the fingerprint features, getting an average error of 0.054 m and a positioning accuracy of 80.4%. This research practiced and discussed various aspects of living wall automatic maintenance systems, providing experimental data and experience to stimulate more ideas and discussions in the future. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/83185 |
DOI: | 10.6342/NTU202203664 |
全文授權: | 同意授權(全球公開) |
顯示於系所單位: | 生物機電工程學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
U0001-2009202217354100.pdf | 4.66 MB | Adobe PDF | 檢視/開啟 |
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。