Vehicle Classification Using Digital Image Processing Based on Convolutional Neural Network for The Electronic Ticketing System in Indonesia

Traffic violations can lead to traffic congestion or even collisions. However, Indonesia’s traffic fining system is still incompatible because officers often take illegal levies. In the development of modern society, a system that can support electronic based to facilitate the community’s needs in the field of transportation, especially in the identification and classification system of vehicle types, is needed. Therefore, this study proposed a system that can classify vehicle types into four classes: motorcycles, cars, trucks, and buses using a Convolutional Neural Network (CNN) method for feature extraction and classification with a self-constructed architecture named NN-Net that consist of three hidden layers and AlexNet architecture which has eight hidden layers. The design model uses various input image sizes, optimizers, and learning rates to get the best model. We achieved an accuracy of 99,14% using the AlexNet architecture with an optimum hyperparameter combination model. Based on our results, the proposed system can be helpful to solve Indonesia’s ticketing system problem.