Development of Longitudinal and Transversal Crack Detection Feature in Automated Road Pavement Condition Survey System Using YOLO Algorithm

Roads are critical infrastructure for economic growth and mobility in Indonesia. To ensure effective maintenance, an application was created to help government conduct road condition surveys automatically using machine learning by detecting road damage. However, the existing one are still insufficient to meet the stakeholder's needs, as it can only detect two types of damages. The aim of this paper is to improve the app's capabilities by increasing the number of damage types that can be detected from previously 2 types (pothole and alligator crack) to 4 types (adding longitudinal and transversal crack). The writers adopts V model system development methodology, a variation of the waterfall model, which consists of stages ranging from requirement modeling to acceptance testing. The result is that the system succeeded in adding two new detection features without reducing the precision of the previous application. The precision obtained is 77% for potholes, 67% for alligator cracks, 64<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup> for longitudinal cracks, and 51 % for transversal cracks. Future work will focus on refining these technologies and expanding the types of detectable pavement damages.