Optical Flow Clustering Using Centroid Neural Network for Motion Tracking of Moving Vehicles

Motion tracking is one of the most practical applications of computer vision in real life. In this paper, the researchers highlight a new application for tracking motion and estimating the velocity of the moving vehicle in terms of clustering of optical flows. A centroid neural network with a metric utilizing optical flow is employed to group pixels of moving vehicles from traffic images, and to generate blobs of moving vehicles. To verify the best optical flow, they utilize Random Sample Consensus by determining the best model that optimally fits the flows. Experiments are performed with various traffic images. The results show that the proposed method can efficiently segment moving vehicles out of background and accurately estimate the velocity of moving vehicle.