Metode Pembentukan Basis Data Spasial Rambu Lalu Lintas dengan Machine Learning

Abstract

Traffic signs are part of road equipment whose existence is very important, in addition to functioning as warnings, prohibitions, orders, or instructions for road users, traffic signs are also a means to reduce accidents and regulate driver behavior. Because of the importance of this function, it is necessary to collect accurate sign data in a spatial database. Currently, there have been many database developments for the management of traffic signs, but data collection is still done manually, by means of surveyors recording groups of traffic signs and entering them into the database. The difficulty faced is the time and accuracy of the surveyors when it comes to selecting groups of signs, this is due to the large number of groups/sub-groups of signs. This problem needs to be solved with the help of a sign group detection tool with an image recognition approach. This study aims to develop an image recognition method to extract photo geotagging information on traffic signs into spatial data and attributes of traffic sign groups. The object of the signs that are sampled are signs that are on roads with the status of provincial roads in the Special Region of Yogyakarta. The results showed that the machine learning-based image recognition accuracy reached 88.66%, further research is needed to improve accuracy by paying attention to the geotagging photo capture variable.