從人工到智慧：建構逢甲大學周邊綠能設施之 GeoAI 動態監測模型
講者/Speaker： 雷祖強, Lei,Tsu-Chiang
傳統都市計畫與土地利用調查極度仰賴人工現地調查或逐棟判讀影像，不僅耗時、費力、成本高，且易因主觀判讀產生誤差，難以進行大範圍的長期動態監測。新世代學生雖熟悉數位工具，但缺乏技術原理與空間資訊整合的實作經驗。本課程旨在「地理資訊系統概論（GIS）」中融入基礎GeoAI概念，將GeoAI影像辨識技術實際導入大學基礎課程中。結合高解析度衛星與航照影像（U-Net），搭配預先訓練好的物件偵測模型，讓學生完整體驗從選區、影像準備、樣本標註到空間統計的GeoAI完整流程。透過物件偵測模型進行大範圍判釋，大幅縮短資料判讀時間，並成功產出GeoAI應用專題成果。
Traditional urban planning and land use surveys heavily rely on manual field investigations or building-by-building image interpretation. This approach is not only time-consuming, labor-intensive, and costly, but also prone to errors due to subjective judgment, making large-scale, long-term dynamic monitoring difficult to achieve. Although the new generation of students is familiar with digital tools, they often lack practical experience in understanding the underlying technical principles and integrating geospatial information. This course aims to integrate foundational GeoAI concepts into the "Introduction to Geographic Information Systems (GIS)" curriculum , actively introducing GeoAI image recognition technology into introductory university courses. By combining high-resolution satellite and aerial imagery with object detection models—such as U-Net —students can fully experience the complete GeoAI workflow, from study area selection and image data preparation to sample labeling and geospatial statistics. Utilizing object detection models for large-scale interpretation drastically reduces image interpretation time and successfully leads to the production of practical GeoAI application project outcomes.
- 論文全文：從人工到智慧：建構逢甲大學周邊綠能設施之 GeoAI 動態監測模型
- 影像： 相關影片