Pemanfaatan Unmanned Aerial Vehicle (Uav) dalam Deteksi Goresan Lamun dengan Pendekatan Machine Learning di Pantai Terora, Tanjung Benoa, Bali

Seagrass propeller; seagrass damage; machine learning; unmanned aerial vehicle (UAV)

Abstract

Seagrass in the Tanjung Benoa area spreads along the eastern part of the beach, including Terora Beach. Tourist activities such as water sports and the movement of fishing boats in the shallow waters of the beach contribute significantly to the damage to the seagrass meadow community. Seagrass scars are formed when boat propellers hit the shallow seagrass meadow floor, destroying the leaves, roots, and rhizomes of the seagrass. The purpose of this study is to determine the level of seagrass scar damage at Terora Beach using the absolute percent scarring formula and to compare the three best machine learningalgorithm for segementing seagrass scars, support vector machine (SVM), random forest (RF), and decision tree (DT). Data collection was carried out in November 2023 using a DJI 4 RTK drone, covering a 350x350 meter area. The machine learning SVM, RF, and DT play an important role in segmenting the level of seagrass damage through the use of unmanned aerial vehicles (UAVs). In addition to the machine learning algorithms, a Median filter is also applied to enhance model evaluation and reduce sand pixels in the middle of seagrass beds in the segmentation results. The SVM evaluation resulted in an Accuracy of 77.19%, Precision of 78.20%, Recall of 77.19%, and F1 Score of 77.30%, which are the highest values compared to the random forest and decision tree algorithms. The median filter will be reapplied, resulting in a map of seagrass scars at Terora Beach. The level of seagrass damage at Terora Beach was found to be 5.835%, categorizing it as moderate scaring.