IDENTIFICATION OF RABAK FLOW (RIP CURRENT) FROM DRONE IMAGE USING CONVOLUTIONAL NEURAL NETWORK (CNN) METHOD IN PALABUHANRATU, SUKABUMI

• Main Author: Fikri Aji Kusuma, Titis
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/50217
• Institution: Institut Teknologi Bandung

Sukabumi Regency is one of the regencies in Indonesia which is endowed with beautiful and attractive coastal areas and is one of the mainstay sectors of tourism that supports the regional economic development of Sukabumi Regency. However, rip currents are encountered on several beaches, namely currents moving toward the high seas at varying speeds which often endanger the lives of tourists. This research will identify rip current images using a machine learning method, namely Convolutional Neural Network (CNN) to classify rip current and the place of occurrence, the data used is 570 images data of the appearance of rip current from the Istana Presiden Beach and Karang Naya Beach, Palabuhanratu. image divided into train and test data with a percentage of 70:30 percent. Three hyperparameter combination scenarios were used including the number of filters in the convolution layer, the number of neurons in the fully connected layer and the learning rate value, which then compared the results to the performance and accuracy value of the model output. Bathymetry and wave data are also used to determine the type of rip current that occurs. The results of the training model show that the model with scenario one can be used to predict the area of rip current occurring at Palabuhanratu Beach, Sukabumi with a model accuracy of 100%, and based on the analysis of oceanographic data, rip current occurring at the Istana Presiden Beach and Karang Naya is a rip current type of topographic rip. .