Detección vehicular mediante teénicas de visión de máquina

Abstract

This paper outlines the results of design of a Haar classifier, which operate according to rectangular descriptors related to the intensity of an image region, for the detection of cars in order to establish the amount of vehicular traffic on a road, supported on the information from video surveillance cameras. The training of the classifier takes place obtaining a percentage of correct detection of 92.9%, and compared the results against machine vision techniques such as optical flow, showing superior performance in more than 30%. Processing times obtained are average of 40 milliseconds.