Tree species mapping using LiDAR and multispectral data based on spectral and object-based image analysis : a case study of Mo Singto in Khao Yai National Park, Thailand | |
| Author | Thantham Khamyai |
| Call Number | AIT Thesis no.RS-20-04 |
| Subject(s) | Optical radar--Thailand--Khao Yai National Park Forest mapping--Remote sensing--Thailand--Khao Yai National Park Multispectral photography--Data processing |
| Note | A thesis submitted in partial fulfillment of the requirements for the degree of Master of Engineering in Remote Sensing and Geographic Information System |
| Publisher | Asian Institute of Technology |
| Abstract | The importance of tree species mapping is to visualize locations of interested trees. Remote sensing knowledge is implemented to measure forest structures in stand community to detect identity of trees in term of physical and chemical properties of tree canopy. Optical sensor on satellite (Sentinel-2A) and LiDAR instrument (high density airborne laser scanning) can observe characteristics of tree identity to distinguish the species of trees. This study aims to extract tree physical structures and chemical composition reflectance to classify tree species using data fusion technique, and map tree species in Mo Singto plot, Khao Yai national park. The models used to classify are Support vector machine, Random forest, Single-layer perceptron and Multi-layer perceptron, and tree species selected to be mapped are Dipterocarpus gracilis, flex chevalieri and Sloanea sigun which are most 3 dominant species in study area. The result showed that MLP and SLP potentially classify more accurate rather than other classifiers. Tree size supposed to be predicted was trees which has DBH more than 25 ern. The overall accuracy was around 0.7 of classification and approximate 0.6 of validation in maps. 4 species maps in this study were created from using only LiDAR metrics and fused LiDAR metrics with spectral indices for each MLP and SLP. Therefore, considered models as selection are 4 model. Additionally, using only spectral indices was not suitable to be used in classification in high complexity tropical forest. Moreover, large size of trees is not good for tree species prediction in every tree size. There is a limitation of segmentation algorithm representing in analysis. Also, medium spatial resolution of multispectral data is not appropriate to operate tree species classification. These issues result in classification to mapping process. |
| Year | 2020 |
| Type | Thesis |
| School | School of Engineering and Technology (SET) |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Remote Sensing (RS) |
| Chairperson(s) | Tripathi, Nitin Kumar; |
| Examination Committee(s) | Sarawut Ninsawat;Sasaki, Nophea; |
| Scholarship Donor(s) | Royal Thai Government Fellowship; |
| Degree | Thesis (M.Eng.) - Asian Institute of Technology, 2020 |