APPLYING UNMANNED AERIAL VEHICLE IMAGES TO INTERPRET MIXTURE FRUIT TREES IN MOUNTAIN AREAS

摘要

Interpreting the crop areas of farmland in Taiwan using remote-sensing technology is particularly difficult, because a variety of crops are planted in a small area, resulting in fragmented and complex farmland patches. The very high-spatial-resolution imagery of unmanned aerial vehicle (UAV) images can be utilized to overcome these difficulties. This study used aerial images taken using UAVs for land cover classification of a study area in Gongguan Township, Miaoli County, followed by the use of image segmentation. The segmentation blocks generated and cadastral maps were combined to perform manual interpretation. The interpretation results indicated that the classification results that underwent multiresolution segmentation were highly effective, with an overall accuracy of 91.6% and a Kappa value reaching approximately 0.9 Due to the low flight altitude, the spatial resolution can be exceedingly high, enabling accurate interpretation of land cover. The results of this study indicate that UAVs are suitable for land cover surveys in small, specific areas.