| Popis: |
With frequent occurrence of natural disasters such as typhoons, and earthquakes annually, Taiwan suffers heavy rains that caused frequent collapses of ridges and mud slides. The objective of this study is to use high-resolution DTM data and their extended geo-morphometric features. Through distinguishing color and geo-morphometric features, the images can be split and merged to form regions. Then, the supervised classification methods, e.g. Support Vector Machine (SVM) and K Nearest Neighbor (KNN) are implemented for the proposed object-oriented analysis. The results show that the producer accuracy (PA) of the SVM and KNN methods are 85.68% and 84.72%, the user accuracy (UA) of the SVM and KNN methods are 80.41% and 79.85%, respectively while applied to the landslide recognition. The SVM offers higher accuracy in recognition mechanism than that of the KNN. The research group plans to continuously explore multiple recognition features and object-driven mechanism to derive the optimum interpretation results. |
