Application of Landsat Multispectral Image and Airborne LiDAR Data on Mountainous Forest Land Cover Mapping

• Main Author: 羅嘉樂
• Other Authors: 林金樹
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/353byd

碩士 === 國立嘉義大學 === 森林暨自然資源學系研究所 === 106 === Multispectral Image Classification can be used to land- cover patterns, but multispectral image classification is susceptible to changes in sunlight and terrain, which are limited by the classification of multispectral images. LiDAR system in the process of access to information will not be affected by changes in solar light and terrain changes, therefore, in this study, we hope to output the relevant information features through the LiDAR system, such as the Canopy Height Model (CHM), texture entropy and gradient to assist the multispectral image of the land cover type classification. Pixel-based and Cluster-Based methods are used in the study. Both classification methods are classified by Support Vector Machine (SVM), and then compare the accuracy of the two classification methods. Selecting the lowest and the highest combination of accuracy in the two classification methods, and then adding the LiDAR feature information for the second classification. The results show that after adding the feature of LiDAR, the overall accuracy of the classification has improved significantly. The pixel-based classification method has the highest accuracy which OA is 89.5% and kappa 0.87. object-base classification method has the highest accuracy which OA is 90.5% and kappa 0.88. The results of this study show that use information of LiDAR characteristics to multispectral image classification has a great help.