Traffic Detection Program using Image Processing and the 1997 Indonesian Highway Capacity Manual (MKJI)

As a part of transport planning and assessment, traffic volume is essential data. However, many Indonesian’s transports engineers, in particular, still use manual traffic counting to collect data series for traffic volume in 24 hours or more. This method is considered inefficient and expensive along with technological developments in this modern era. The purpose of this study tries to simplify the traffic counting method by developing an automatic vehicle detection program based on the 1997 Indonesian Highway Capacity Manual (MKJI) classification. The application that used in this study is KS Traffic Analyzer as open-source code traffic counting based on Java and OpenCV library. This application then modifies as MKJI 1997 vehicles classification to measure traffic data in urban and rural road using drone quadcopter. This tool using Gaussian Mixture Model (GMM) method for processing images/videos of traffic based on the background and foreground. Traffic footage was tested at different times, height, and angle of shooting. The accuracy was measured by comparing the volume from the program and manual counting. The results showed that the best accuracy between real traffic volume and automatic counting in the program is in the urban street with the best accuracy reached 93.66% which was taken video in the morning and the height is 5 m. This result of this study also answers the function of road performance based on volume per capacity.