使用半自動變異偵測方法偵測地表覆蓋變化;Using A Semi-Automatic Change Detection Algorithm to Detect Land Cover Changes

NCUIR > Research Center > Master of Science Program in Remote Sensing Science and Technology > Electronic Thesis & Dissertation > Item 987654321/93423

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/93423

Title:	使用半自動變異偵測方法偵測地表覆蓋變化;Using A Semi-Automatic Change Detection Algorithm to Detect Land Cover Changes
Authors:	李品萱;LI, PIN-HSUAN
Contributors:	遙測科技碩士學位學程
Keywords:	變化檢測;影像分割;多變異分析;SPOT-7;Pléiades;Change Detection;Image segmentation;Multivariate Analysis;SPOT-7;Pléiades
Date:	2023-07-31
Issue Date:	2024-09-19 17:01:33 (UTC+8)
Publisher:	國立中央大學
Abstract:	地表覆蓋的變化在人為活動及自然作用下是動態且無預警的發生，特別是在自然災害及戰爭事件之緊急事件發生時，其需要一套工作流程來獲取變化的資訊，隨著遙測技術越漸成熟，硬體設備的提升及高解析度的影像容易取得，讓遙測技術可以做為偵測地表覆蓋變化之工具。然而，在現有物件式變化檢測方法進行影像分割時，需要透過調整分割方法的參數及視覺化確認分割圖層的正確性，在實際應用層面中較需要花費時間達成，並且在變化檢測方面使用如人工智慧等工具時，變化檢測的過程不是透明化，無法得知變化特徵的重要性。因此，本研究嘗試發展一個半自動變化檢測之方法，以改善以上不足之處，此方法將基於物件式之變化檢測方法繪製出地表變化覆蓋圖，在嘉義的日常地表覆蓋變化檢測中，透過 SPOT-7 影像找出突然變化之區域，及在花蓮的 918 地震事件中應用 Pléiades影像偵測損毀的橋樑及季節性地表覆蓋。與此同時，我們會在這兩個案例中對影像在相對輻射校正及特徵篩選的變化檢測、影像在未相對輻射校正及未特徵篩選的變化檢測，及未相對輻射校正及使用特徵篩選的變化檢測進行精度評估。變化檢測的結果在嘉義的案例中可以找到農地覆蓋的變化外，也試圖找出不是季節性變化之區域，最高的總體精度為 86%。而在花蓮案例的總體精度最高達到84%，在花蓮的變化圖中，除了繪製出季節性地變化之外，也繪製出因地震而導致損毀的高寮大橋。在三個地表變化偵測的測試中，比較特別的是影像在有相對輻射校正後的變化檢測精度比未校正影像的變化檢測的成效不佳，其原因與相對輻射校正的方法有所關聯。;The changes in land cover occur dynamically and unpredictably due to both human activities and natural processes. Particularly during emergencies such as natural disasters and war events, there is a need for a change detection (CD) workflow to obtain information on land cover changes. With the advancement of remote sensing technology, improved hardware, and easy accessibility to high-resolution imagery, remote sensing has become a valuable tool for detecting land cover changes. However, existing object-based CD methods face challenges in practical applications, it is necessary to spend time adjusting the parameters of the segmentation method and visually analyzing the accuracy of the segmentation layer. This process requires more time to achieve in practical application scenarios. For those algorithms utilizing tools such as artificial intelligence (AI) lack transparency, making it difficult to assess the importance of change features. Therefore, this study aims to develop a semi-automatic change detection method to address these limitations. This method utilizes object-based CD to generate land surface change maps. In the case of abrupt change areas in Chiayi, sudden change areas are identified using SPOT-7 imagery. In the case of the 918 earthquake event in Hualien, Pléiades imagery is used to detect damaged bridges and seasonal land surface cover changes. In both cases, an accuracy assessment is conducted for CD with relative radiometric calibration (RRC) and feature screening (FS), CD without relative radiometric calibration and feature screening, and CD without relative radiometric calibration but with feature screening. The detection results in the Chiayi case reveal not only changes in agricultural land cover but also attempts to identify non-seasonal change areas, with a maximum overall accuracy of 86%. In the Hualien case, the overall accuracy reaches up to 84%, and the change map includes both seasonal variations and the collapsed Gao-Liao Bridge caused by the earthquake. Among the three tests of our results, it is noteworthy that the CD accuracy after atmospheric correction is less effective compared to CD without relative radiometric calibration, which is associated with the method of atmospheric correction.
Appears in Collections:	[Master of Science Program in Remote Sensing Science and Technology ] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	22	View/Open

社群 sharing

Loading...