Object based classification of unmanned aerial vehicle (UAV) imagery for forest fires monitoring

In case of fire, determination of burned trees and fire direction is very important. Until now, satellite images and aerial photographs have been widely used in forest fire studies. However, these data can be ineffective in terms of temporal and spatial resolution. In recent years, due to high resolution of provided images, use of Unmanned Aerial Vehicle (UAV) rapidly increased in forest related monitoring studies. Images obtained soon after the forest fires by means of UAVbecome the most important data to evaluate the damage level in the forestry area using different classification techniques. Conventional image classification methods are inefficient for evaluation of high resolution images. However, object-based classification is more accurate than conventional methods. Because, this method uses spectral, neighborhood, texture, hierarchy and size based relationships. In this study, forest fires occurred in Camburnu Natural Park in Surmene District of Trabzon Province located in the Black Sea Region of Turkey was selected as the study area. To determine the destroyed area, high resolution UAV images of the study area were obtained and image pre-processing steps were employed. Object-based classification and pixel-based classification have applied to these images. The boundaries of destroyed forest have been extracted by means of two classification methods. Additionally, combining of these two classification results was investigated to improve the results of the burned area.