| DC 欄位 |
值 |
語言 |
| DC.contributor | 資訊工程學系 | zh_TW |
| DC.creator | 黃家茹 | zh_TW |
| DC.creator | Chia-Ju Huang | en_US |
| dc.date.accessioned | 2023-7-25T07:39:07Z | |
| dc.date.available | 2023-7-25T07:39:07Z | |
| dc.date.issued | 2023 | |
| dc.identifier.uri | http://ir.lib.ncu.edu.tw:88/thesis/view_etd.asp?URN=110522121 | |
| dc.contributor.department | 資訊工程學系 | zh_TW |
| DC.description | 國立中央大學 | zh_TW |
| DC.description | National Central University | en_US |
| dc.description.abstract | 本研究旨在進行台灣地區PM2.5濃度的預測和補值，並與其他概念做出明顯區別。相較於傳統方法，我們將PM2.5濃度視為圖像，並運用圖像修補技術進行缺失值的修復。此外，我們特別強調僅使用經緯度地理資訊和PM2.5濃度作為輸入特徵，而不考慮其他氣象特徵。
為了實現此目標，我們收集了台灣各地的大量PM2.5濃度數據和相對應的經緯度地理資訊。首先，我們將PM2.5濃度數據轉換為圖像表示，其中每個像素點代表一個觀測站點的PM2.5值。然後，利用圖像修補技術，我們根據周圍已有的PM2.5觀測站點數據，預測並填補目標區域的缺失值。這種基於圖像的方法使得我們能夠捕捉到空間鄰近性和相關性，從而改善缺失值的補值效果，透過將PM2.5濃度視為圖像並運用圖像修補技術，與傳統方法有明顯的區別。。
為了驗證我們方法的有效性，我們進行了一系列實驗和比較。結果顯示，我們提出的基於圖像修補的方法在PM2.5濃度的預測和補值方面具有潛力。此外，我們的方法利用了僅使用經緯度地理資訊和PM2.5濃度作為輸入特徵的特點，使其在資料需求和計算複雜度方面相對簡化，這項研究的成果有望為空氣污染監測和環境保護提供有價值的參考和指導。 | zh_TW |
| dc.description.abstract | This study aims to predict and interpolate PM2.5 concentrations in Taiwan while distinguishing itself from other approaches. In comparison to traditional methods, we treat PM2.5 concentrations as images and utilize image inpainting techniques for missing value restoration. Additionally, we specifically emphasize the use of only geographical information (latitude and longitude) and PM2.5 concentrations as input features, excluding other meteorological factors.
To achieve this objective, we collected a substantial amount of PM2.5 concentration data and corresponding geographical information (latitude and longitude) from various locations in Taiwan. Firstly, we transformed the PM2.5 concentration data into image representations, where each pixel represents the PM2.5 value at an observation station. Then, using image inpainting techniques, we predicted and filled in the missing values in the target areas based on the surrounding PM2.5 observation station data. This image-based approach allows us to capture spatial proximity and correlations, thereby improving the effectiveness of missing value interpolation. By treating PM2.5 concentrations as images and applying image inpainting techniques, our approach distinguishes itself from traditional methods.
To validate the effectiveness of our method, we conducted a series of experiments and comparisons. The results demonstrate the potential of our proposed image inpainting-based method in predicting and interpolating PM2.5 concentrations. Furthermore, our method′s reliance solely on geographical information (latitude and longitude) and PM2.5 concentrations as input features simplifies data requirements and computational complexity. The outcomes of this research are expected to provide valuable references and guidance for air pollution monitoring and environmental protection. | en_US |
| DC.subject | PM2.5 | zh_TW |
| DC.subject | 空氣污染 | zh_TW |
| DC.subject | 插值 | zh_TW |
| DC.subject | 圖像修補 | zh_TW |
| DC.subject | 缺失值修復 | zh_TW |
| DC.subject | PM2.5 | en_US |
| DC.subject | Air pollution | en_US |
| DC.subject | Interpolation | en_US |
| DC.subject | Image inpainting | en_US |
| DC.subject | Missing value restoration | en_US |
| DC.title | 透過圖像修補技術補空間中的PM2.5值：以台灣為例 | zh_TW |
| dc.language.iso | zh-TW | zh-TW |
| DC.title | Imputing spatial PM2.5 values via image inpainting: A case study in Taiwan | en_US |
| DC.type | 博碩士論文 | zh_TW |
| DC.type | thesis | en_US |
| DC.publisher | National Central University | en_US |