Sistema de detecção de veículos e avaliação de eixos por meio de máquina de vetor de suporte.

Abstract

This work aims to develop a software for the detection and evaluation of vehicle axles through the use of a support vector machine classifier. With its use, we intend to increase the agility of industrial processes related to counting and evaluating vehicle axles that travel through a toll that was chosen as a case study, near the city of Santa Cruz do Sul. This research is characterized by being of academic level, exploratory regarding the bibliographic survey and technical procedures. It is also a case study, because it uses video footages as a way of sampling and data analysis. For the implementation of the proposed software, the Python programming language was used in conjunction with the Open Source Computer Vision Library (OPENCV). Among the image processing techniques used, we highlight Gaussian Blur, background definition, threshold, contour detection, Hough transform and histogram of oriented gradients. The classifier support vector machine was employed to validate the detected objects, assuring that they refer to a vehicle axle. Tests using a road traffic footage captured by the authors indicate that 97% of cars and 52% of trucks had their axles correctly accounted by the developed software.