Detection and Classification of Road Damage Based on Water-Filled Pothole Using Convolutional Neural Network Model for Edge Device

Road damage surveys in Indonesia are still conducted manually through visual inspections based on the Surface Distress Index (SDI) method. Consequently, the process often requires extended completion times and yields results that lack objectivity due to heavy reliance on the surveyor’s experience. As a result, road repairs frequently do not correspond accurately to the actual damage conditions. Road deterioration intensifies during the rainy season, when water accumulates in potholes, accelerating their erosion and expansion. To facilitate more objective damage assessment, particularly for potholes, a tool employing an image sensor capable of distinguishing between water-filled and dry potholes is necessary. This study utilized an image processing model based on a convolutional neural network employing MobileNet SSD V2. In detecting water-filled potholes, the system achieved a precision of 0.95, a recall of 0.514, and an F1 score of 0.667. Furthermore, performance testing across various vehicle speeds indicated that the optimal speed for the edge device system was an average of 15 km/h, at which the system maintained a precision of 0.95, a recall of 0.514, and an F1 score of 0.667.