Please use this identifier to cite or link to this item:
標題: 非監督式高頻譜資料開發
作者: 張建禕
關鍵字: Hyperspectral imaging;高頻譜影像;資訊工程硬體工程;Automatic Target Recognition (ATR);Unsupervised Linear Spectral Unmixing (ULSU);Real-Time Data Processing;非監督式訊號處理;自動的目標識別;分類;非空間訊號處理;技術發展
Hyperspectral imaging has become an emerging technique in remote sensing community, but it is still in its infancy in Taiwan due to the lack of understanding its potential in applications. Accordingly, one of its major goals is to facilitate research in hyperspectral imaging in Taiwan. Another is to present in professional societies and publish papers in peer reviewed journals. A third goal is to promote visibility of National Chung Hsing University in remote sensing community. A fourth goal is to establish working relationship between National Chung Hsing University and government agencies of agriculture, forest, law enforcement, national security, medical imaging etc and explore potential projects. Last but not least is to plant a seed for a possible setup for a remote sensing laboratory or center for excellence in national Chung Hsing University. With recent advances of hyperspecral remote sensing technology the utility of hyperspectral imagery covers a wide range of military and civilian applications. Of particular interest are targets with their small presence and low probability existence in data exploitation. Such targets include special spices in agriculture and ecology, toxic wastes in environmental monitoring, rare minerals in geology, drug/smuggler traffic in law enforcement, military vehicles and landmines in large battlefields, chemical/biological agents in bioterrorism and weapon concealment and mass graves in intelligence gathering. All of these targets are generally considered as insignificant targets and cannot be found by traditional spatial domain-based image processing, but are indeed of major interest from an intelligence point of view. The great challenge in detection and recognition of such targets is that these targets provide very limited spatial information and are generally difficult to visualize in data. Hyperspectral imaging offers an effective means to detect, discriminate, classify, quantify and identify such targets using their spectral characteristics captured by high spectral-resolution sensors without accounting for their spatial information. The processing techniques that only make use of spectral properties without taking into account spatial information is referred to as non-literal (spectral) processing techniques in hyperspectral imaging. Therefore, one of great interests in hyperspectral sensing is to develop unconventional non-literal processing techniques for hyperspectral data exploitation. Six tasks, Task 1: Estimation for Number of Signal Sources. Task 2: Automatic Target Extraction (ATE). Task 3: Automatic Target Recognition (ATR). Task 4: Unsupervised Linear Spectral Unmixing (ULSU). Task 5: Data Visualization, Task 6: On-Board/Real-Time Data Processing, are proposed in the project to address several long standing key issues but yet to be resolved in hyperspectral imaging. They are fundamental data processing steps for unsupervised non-literal analysis to be effective for hyperspectral data exploitation.

高頻譜影像在遙測領域中已經成為一種先進的技術,由於在台灣缺乏潛在的應用,目前仍然是屬於初期發展的階段。因此,第一個主要的目標是促進在台灣高頻譜影像分析的研究。另一個藉此把研究成果發表在專業領域的論文在學術期刊上。 第三個目標是提升國立中興大學在遙測影像分析領域中的能見度。第四個目標是將建立國立中興大學和政府及學術機構的合作關係,例如:森林族群的分類、犯罪偵測、國家安全、醫學影像等等。最後是對於國立中興大學在未來成立遙測影像技術實驗室或卓越遙測中心鋪路。由於高頻譜影像遙測技術已漸趨成熟,高頻譜影像實際應用上包括廣泛的軍事和民間應用。其中特別感興趣的是在資料中搜索屬於小而不易發現的目標。一般而言低出現機率的目標物。這類的目標常出現在農業和生態學方面,例如:在環境監控過程中的有毒廢棄物、在地質學方面的稀有礦物質、在毒品/走私、軍用車輛和地雷、生物病毒、武器隱藏情報收集。這些目標大多不明顯而不易被發現,且不能被以傳統影像處理技術所發現,但是從情報的觀點來看,這些訊息反而最重要而且是我們最感興趣的地方。但是在偵測與識別這些目標卻是一個重大的挑戰,主要因為這些目標所能提供的空間訊息非常有限並且一般難用數據表現。高頻譜影像則提供一個在偵測、區別、分類、量化、認證目標物有效方法。更重要的是高頻譜影像處理技術利用頻譜的特性來分析數據而不需要解釋數據的空間訊息。這類利用光譜的特性而沒有考慮到空間訊息處理的技術稱為非空間高頻譜影像處理。在這個計畫中我們提出了六項目標。目標一︰非監督式的訊號處理、目標二︰自動的擷取目標物、目標三︰自動的目標識別、目標四︰非監督式的線性光譜分層、目標五︰資料視覺化、目標六︰即時的訊號處理,這些重要的資料處理的步驟對於非監督式、非空間訊號處理的分析在高頻譜影像處理是非常有效的。
其他識別: NSC98-2221-E005-096
Appears in Collections:電機工程學系所

Show full item record

Google ScholarTM


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.