| Summary: | 碩士 === 國立成功大學 === 資源工程學系碩博士班 === 96 === Due to the rapid development of remote sensing technology, the remote sensed image data had widely used in evaluating the forest resource inventory. Traditional method to estimate the NPP with AVHRR or MODIS satellite images. The penetrating rate of plantation did not considered in such process. Therefore, only the NPP from canopy of tree is estimated.In this study, the object-oriented classification process consists of two various steps: image segmentation and object classification. In the image segmentation step, the process group the pixels into numerous sub-regions thus transform the source image into an objective image. The second step is designed to classify every sub-region according to the database of experts. This classification is based on knowledge-based system which was designed by spectrum and texture information. LiDAR is capable of measuring objects with very high density and multiple return signal among tree crowns, therefore it can rapidly obtain the 3D data about forest stands. Capability ofpenetrating over forestry is a primary factor for this techniques being recognized in recent years.The proposed schema in the study comprises four major steps: (1) LiDAR data preprocessing, (2) Object-Oriented Classification by vegetated regional extraction, (3) LiDAR data in vegetated regional penetration rate calculation. (4) Estimate NPP by FORMOSAT-2 images and LiDAR data.There are two study areas, mountains and plains, in the study. Research results show that the NPP estimated by LiDAR is other than NPP calculated from FORMOSAT-2 images. In the plains area +10.7278% more and in the mountains area +9.7889% increased value was found. Since the vegetation in plains region is sparser than in mountains, according to the data of this research, penetrating rate 17.50848% in plains, while the same phenomena of 15.87155% found in mountain area. The calculated NPP at plains is larger then region of mountain.
|
|---|
