跳到主要內容 :::

年度	109	專案性質	實驗性質
專案類別	研究專案	研究主題	調查
申請機構	東海大學	申請系所	環境科學與工程系
專案主持人	陳鶴文	職等/職稱	教授
專案中文名稱	無人載具與高光譜技術在土壤重金屬監測上的應用
中文關鍵字	無人載具,高光譜,土壤重金屬,卷積神經網路
專案英文名稱	Soil Pollution Monitoring by Application of Unmanned Aerial Vehicle and Hyperspectral Imaging
英文關鍵字	Unmanned Aerial Vehicle, Soil Pollution Monitoring, Convolutional Neural Network
執行金額	830,000元
執行期間	2019/12/1 至 2020/11/30
計畫中文摘要	環保署土基會積極針對於全國農地進行重金屬污染調查，藉以確保農糧生產安全，將高污染潛勢農地進行採樣，建立污染潛勢農地管制與整治工作，並透過過去土壤背景資料，建立綜合指標評價系統。而此系統因台灣之農地重金屬高污染潛勢地區幅員廣大，需耗費大量人力、物力及經費，才能掌控整個區域中實際污染情形，因此建立即時重金屬監測模型具有其必要性。為此，本團隊將建置無人飛行載具、高光譜影像及重金屬智能辨識之整合型技術，階段內容為：(1)利用農地之無人載具高光譜影像設備建置；(2)高光譜波段處理技術；(3)應用人工智慧之卷積神經網路進行農地重金屬辨識模型，此技術完成後，本技術可建置該地區的土壤污染趨勢圖，並可快速提供政府機關判斷農地是否列管或進入下一階段管制之參考依據，以提升政府機關調查污染潛勢之效率及減少相對應之成本支出。
計畫英文摘要	The Soiland Groundwater Pollution Fund Management Board has been conducting the national heavy metal pollution surveys on agricultural lands to ensure the safety of agricultural production. It analyses the soil samples taken from the agricultural lands with high potential of contamination, and combines the historical soil data to establish an indicator database. However, it requires huge amount of manpower and budgets to investigate the pollution situation in the whole country. Therefore, build an immediate heavy metal monitoring model is crucial. The team plans to build an integrated technology combining drone, hyperspectral imaging instruments, and artificial intelligence techniques, and follow the three steps to build the model: (1) taking hyperspectral images of agricultural lands with drone, (2) hyperspectral image processing, and (3) Identify heavy-metal polluted lands with convolutional neural network (CNN). The model will then be used to map the soil pollution trend in the region, and provide effective information to the government in estimation of the heavy-metal pollution indicator of the agricultural lands. It will improve the efficiency of government agencies in investigating pollution potential and reduce the corresponding cost.