Date of Publication
8-12-2024
Document Type
Bachelor's Thesis
Degree Name
Bachelor of Science in Physics with Specialization in Materials Science
Subject Categories
Physics
College
College of Science
Department/Unit
Physics
Thesis Advisor
Edgar A. Vallar
Defense Panel Chair
Maria Cecilia D. Galvez
Defense Panel Member
Ofelia T. Rempillo
Prane Mariel P. Ong
Abstract (English)
The environmental effects of anthropogenic climate change have had drastic implications for the health of many coastal ecosystems in the Philippines. Despite this, recent innovations in the technologies of UAVs, multispectral sensors, and machine learning provide a possible solution to this problem. This study proposes the use of a multispectral UAV-machine learning-based methodology for the rapid characterization of these environments. The study was conducted in 2023 over a coastal region near Lian, Batangas. A multispectral DJI Mavic 3M UAV was used to image the region with an RGB and a G/R/RE/NIR multispectral camera. This data was then compiled into NDVI, GNDVI, NDRE, OSAVI, and LCI maps of the region. OSAVI was deemed unsuitable for this task. Together with manual labeling, the k-means clustering algorithm was then used to identify ecotopes present within the region to be used as training data. The generated training data was then used to train SVM and Random Forest models for pixel-based ecotope classification using a 5-folds cross-validation set up. This resulted in an SVM model suitable for microprocessor devices with an accuracy of 0.90 and a RF model suitable for highly accurate post-processing with an accuracy of 0.99. These models were then used to characterize the entire image in comparison to the training data. It was found that though there existed a statistically significant difference in the vegetation index values, this difference had an expectation value of ±0.05 and ±0.002 for the SVM and RF models respectively.
Abstract Format
html
Abstract (Filipino)
"-"
Abstract Format
html
Language
English
Format
Electronic
Keywords
Drone aircraft; Multispectral imaging
Recommended Citation
Pe, S. T. (2024). Classification of coastal ecotopes through the use of UAV imaging and machine learning methods. Retrieved from https://animorepository.dlsu.edu.ph/etdb_physics/36
Upload Full Text
wf_yes
Embargo Period
9-15-2027
