Detecting and Counting the Moving Vehicles Using Mask R-CNN

In order to make use of the existing video surveillance facilities and improve the ability of urban intelligent traffic integrated management, a novel moving vehicle detecting and tracking method based on traffic video is proposed in this paper. The Mask R-CNN algorithm based on deep learning is used to detect vehicle contours in the complex urban traffic environment, and then the improved Kalman filter is applied to track the vehicle target in the video sequence. The experimental results show that the proposed method can reach 95% average accuracy with 2.86 fps speed, and can effectively deal with different traffic and climates conditions.