A Deep Learning Approach for Traffic Incident Detection in Urban Networks

Abstract

Incident detection function is vital for traffic control and management and is an important prerequisite for quick restoration of smooth traffic flow in urban networks. With accurate and reliable incident detection, a wide range of environmental and economic benefits can be realised by mitigating congestion quickly. This paper proposes a deep learning method, Convolutional Neural Networks (CNN), for automatic detection of traffic incidents in urban networks by using traffic flow data. The method was evaluated using traffic flow and incident datasets from Central London. Performance indexes, such as False Positive Rate, Detection Rate, Precision and F-measurement, are used to comprehensively evaluate the performance of the proposed method in comparison with a conventional machine learning method, i.e., Multi-Layer Perceptron. The results demonstrate that the proposed method may be superior to traditional neural networks with a higher Detection Rate and a lower False Positive Rate. The results also indicate that deep learning-based incident detection may improve the accuracy of incident detection, especially in a large urban network.