Implementation of deep learning based semantic segmentation method to determine vegetation density

The dryness of peatlands is influenced by the density vegetation. If are dry, they become vulnerable to a fire risk. To calculate drought index, professionals must conduct vegetation analysis. However, field analysis requires vast amounts resources. Moreover, accuracy based on satellite data not adequate. Therefore, this research presents drone-captured two-dimensional image data. object Liang Anggang Protection Forest Block I in Banjarbaru, South Kalimantan, Indonesia. It surveyed for information its cover. Afterwards, There 300 images cover collected and utilized total. method deep learning with semantic segmentation will be used compare results determining methods expert as ground truth. contribution study determine optimal performance model classifying into three categories: bare/ungrazed, lightly grazed, heavily grazed. Performance evaluated correctness intersection over union (IoU). Obtaining proper parameters classification using techniques comparing best objectives following contribution. From experimental studies conducted, momentum parameter value MobileNetV2, Xception, Inception-ResNet-v2 0.9, 82.69 percent average. most appropriate ResNet 18 architecture 0.1. result DeepLabV3 estimating compared U-Net model.