The identification and estimation of health of coconut trees by using Unmanned Aerial Vehicles (UAV) data and a machine learning method | |
| Author | Panruthai Tangprasert |
| Call Number | AIT Thesis no.RS-17-28 |
| Subject(s) | Coconut-palm--Remote--Sensing Machine learning |
| Note | A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Remote Sensing and Geographic Information Systems |
| Publisher | Asian Institute of Technology |
| Series Statement | Thesis;no. RS-17-28 |
| Abstract | Coconut trees have been damaged by the epidemic coconut insect pests for long years, but the current survey cannot allow an efficient management and control because this method does not support timely data. In this work, an alternative method is presented for collective the coconut trees information, which indicated the technique to automatically identify coconut tree and estimates health of coconut trees by using individual UAV images combined with machine learning method. Moreover, the properties of individual UAV images are described their effects on automatic identification. The automatic identification number of coconut trees is developed by applying Histogram of Oriented Gradients (HOG) and Support Vector Machine (SVM), which presents 76.29%, 58.69% and 65.25% of the average of precision, recall, and F1 - score respectively, and the properties of UAV images are shown, which has the relationship between the performance of identification results and the properties of UAV image input, which found many factors that affect the quality of detection, including 3 triangles of exposures and weather condition. The recommendations to create the suitable UAV images for identification are a small aperture to reduce overbright on images, a high speed to decrease blur on images, and a low ISO to avoid noise on images, and weak wind and enough light condition are also suggested when UAV flights. In part of health estimation by colors of the canopy, the best performance of classifier provides the average of precision of 69.69%, recall of 64.80%, and F1 - score of 61.47% from Logistic Regression. This alternative method has a potentiality for collective preliminary coconut information and reduces cost - expensive, labor - intensive and time - consuming, which can promote frequency survey, which supports efficient coconut management and surveillance of the epidemic coconut insect pests. |
| Year | 2017 |
| Corresponding Series Added Entry | Asian Institute of Technology. Thesis;no. RS-17-28 |
| 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;Masahiko Nagai |
| Examination Committee(s) | Miyazaki, Hiroyuki;Dailey, Mathew N. |
| Scholarship Donor(s) | Royal Thai Government Fellowship |
| Degree | Thesis (M.Sc.) - Asian Institute of Technology, 2017 |