۱- Introduction

۲- Convolutional Neural Network

۳- Research Method

۴- Results and Analysis

۵- Conclusion and Future Work

References

Abstract

In this paper, we present an intelligent traffic congestion detection method using image classification approach on CCTV camera image feeds. We use a deep learning architecture, convolutional neural network (CNN) which is currently the state-of-the art for image processing method. We only do minimal image preprocessing steps on the small size image, where the conventional methods require a high quality, handcrafted features need to do manual calculation. The CNN model is trained to do binary classification about road traffic condition using 1000 CCTV monitoring image feeds with balance distribution. The result shows that a CNN with simple, basic architecture that trained on small grayscale images has an average classification accuracy of 89.50%.

Introduction

Road traffic condition is one of major problems in big cities especially in the developing countries, like in Indonesia. There are so many bad effects caused by this traffic problem, like massive delays and the increased fuel wastage and monetary losses [1]. The road traffic could be happened because of limited number of facilities and infrastructures, huge number of people, and inappropriate policy from the government. Based on the data from tomtom traffic index [2], Jakarta is the 3rd city with traffic condition in the world. In order to tackle this problem, Indonesian government, especially Jakarta’s government try to implement the concept of smart city technology. The government provides CCTV in selected area to monitor the traffic condition. Intelligent Traffic System (ITS) has been developed to solve the road traffic condition. It uses supported data, such as airbone optical remote sensing sensor [3], wireless signal communication (i.e probe vehicle-to-vehicle) [4][5][6][7]. Nevertheless, in the developing countries, the first problem is that the data is not available because of the expensive infrastructure and maintenance cost. Another alternative is utilizing data from traffic video [8] and captured image [1] with manually processing by human. Manually processing requires handcrafted features and manual calculation like calculating the distance and the level of the traffic congestion between vehicle. Manually processing depends on the human ability and requires time which is not short. Therefore, in this research, we propose a method to detect the road traffic congestion automaticly by utilizing the data from CCTV camera image feeds. We conduct a series of computaional experiments for the road traffic data in Jakarta. We implement the Convolutional Neural Network and preprocessing of the data.