Land Use Classification Technique Using Unmanned Aerial Vehicle Orthophotos

Main Article Content

Kraiwut Siri-on
Piyawat Diloksumpun
Weeraphart Khunrattanasiri

Abstract

Unmanned Aerial Vehicle (UAV) can currently take high-resolution orthophotos but the camera sensor only has an RGB band. Classifying land use using UAV aerial photos through computer needs suitable techniques. The objective of this research was to find suitable techniques for such land use classification. Pixel and object-based classification techniques were applied and validated using visual interpretation techniques. The study area was in the Ban Pa Not, Mae Tha subdistrict, Mae On district, Chiang Mai province, in a total area of approximately 1,182.47 rai, with some parts in the Pa Khun Mae Tha National Reserved Forest.


We found that pixel-based classification techniques have an overall accuracy of 39.4% with a Kappa coefficient of 0.27, while object-based classification technique has an overall accuracy of 34.2% with a Kappa coefficient of 0.22.  Pixel-based classification techniques gave a slightly higher value but both techniques had an acceptable level of performance as indicated by the Kappa coefficient values on the Landis and Koch scale. We conclude that orthophotos in the RGB band obtained from UAVs are not suitable for land use classification and it is nescessary to find additional techniques to validate such classifications.


 

Article Details

How to Cite
Siri-on, K., Diloksumpun, P., & Khunrattanasiri, W. (2021). Land Use Classification Technique Using Unmanned Aerial Vehicle Orthophotos. Thai Journal of Forestry, 40(2), 95–110. Retrieved from https://li01.tci-thaijo.org/index.php/tjf/article/view/249908
Section
Original Articles