Vehicle classification and speed estimation using Computer Vision techniques

Abstract

In this work, we implement a real-time vehicle classification and speed estimation system and apply it to videos acquired from traffic cameras installed in highways. In this approach we: a) Detect moving vehicles through backgroundforeground segmentation techniques. b) Compare different supervised classifiers (e.g. artificial neural networks) for vehicle classification into categories: (car, motorcycle, van, and bus/truck). c) Apply a calibration method to georeference vehicles using satellite images. d) Estimate vehicles speed per class using feature tracking and nearest neighbors algorithms.