Oil palm tree detection and health classification on high-resolution imagery using deep learning technique | |
| Author | Kanitta Yarak |
| Call Number | AIT Thesis no.RS-19-20 |
| Subject(s) | Deep learning (Machine learning) Palm oil--Health aspects Oil palm |
| Note | A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Remote Sensing and Geographic Information Systems, School of Engineering and Technology |
| Publisher | Asian Institute of Technology |
| Abstract | Applying modem technology together with agriculture is an important way which can result in effective management. The technology used in this study is an alternative way for oil palm tree management by applying high-resolution imagery cooperate with Faster-RCNN for automatically detection and healthy classification. This research uses a total of 4,172 bounding boxes of healthy and unhealthy palm trees which were built from images of 2000 by 2000 pixels. The 90 percent of a total dataset is used as training and 10 percent is prepared for testing on Resnet-50 and VGG-16 model. The results of this research present that there are three important significant for oil palm tree detection and healthy classification using our method which are the crown size, the crown color, and the crown density. For automatically detection and health classification, the Resnet-50 reach the higher classifier accurate after training in 98.22%. In case of test results, a Resnet-50 has a better percentage of the Fl-score on oil palm tree class, healthy oil palm tree class, and unhealthy oil palms in 95.09%, 92.07%, and 86.96% respectively when compared with visual interpretation of UA V imagery. Consideration of the model performance after comparing Resnet-50 results with ground sampling shows that the F l-score of oil palm tree class, healthy oil palm tree class, and unhealthy oil palms tree class has a percentage in 97.67%, 95.30%, and 57.14.% respectively. Based on our study results, it can be concluded that training and testing models with the same dataset, the Resnet-50 network has better performance on detection and health classification. |
| Year | 2019 |
| Type | Thesis |
| School | School of Engineering and Technology |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Remote Sensing (RS) |
| Chairperson(s) | Apichon Witayangkurn |
| Examination Committee(s) | Miyazaki, Hiroyuki ; Sarawut Ninsawat |
| Scholarship Donor(s) | His Majesty the King’s Scholarships (Thailand) |
| Degree | Thesis (M. Sc.) - Asian Institute of Technology, 2019 |