Deep Learning Based Car Detection System Using Convolutional Neural Network and Haar Cascade Classifier

Advances in technology have played an important role in increasing the efficiency of the transportation system in Indonesia. However, the rapid growth in private vehicles has also caused significant traffic congestion in many areas. An innovative solution is needed to manage traffic, and one of the proposed solutions is using a Smart Traffic Light System. This system utilizes the latest vehicle object detection technology to dynamically adjust the duration of traffic lights based on the number of vehicles detected in each lane. This study presents an approach using two vehicle detection methods: the Convolutional Neural Network (CNN) and the Haar Cascade Classifier. The average accuracy value generated by the system is 86%, but this method shows a very high level of precision, reaching 100%, which means there are almost no errors in detecting vehicles. The recall of this method also reaches 86%, indicating the ability to detect most of the vehicles passing through the traffic lane. The results of this study indicate a strong potential for implementing a vehicle detection system with a combination method of CNN and Haar Cascade Classifier to overcome traffic congestion problems. The use of sophisticated and intelligent detection technology in the Smart Traffic Light system is expected to increase traffic efficiency, reduce waiting times, and optimize road use. Thus, this research makes an essential contribution to the development of a transportation system that is more adaptive and responsive to dynamic traffic.