Online vehicle detection using gated recurrent units

As road transportation increases and inner and outer city roads are extended, traffic control systems and intelligent monitoring of traffic flow are crucial. Vehicle detection is one of the primary tasks in computer vision applications, which is widely used in surveillance-related tasks, especially in intelligent traffic monitoring systems. To address this issue, a number of approaches have been presented out, the most of which are based on deep learning frameworks. We propose a method that solves the vehicle detection problem by online detection without needing labeled data so it can be used for traffic flow supervision. The proposed approach utilizes the deep gated recurrent neural networks that can extract moving objects (vehicles) from the original scenes by applying background model maintenance. The method is evaluated on BMC database of traffic videos. As the evaluation process is carried out from the first frame to the end of the video, the experimental results are presented by the well-known metrics and visually demonstrated, which is remarkable considering the unsupervised training and real-time performance.

Keywords:

Vehicle detection , Background model maintenance , Deep learning , Gated recurrent unit , Online training , Traffic surveillance systems.