Testes de desempenho de algoritmos de deteção de artefatos em Plataforma Embarcada Raspberry Pi Utilizando OpenCV e Node.JS.

Abstract

This work of completion of course contemplates the development of a system of experi- mentation of three methods of feature matching: ORB, SIFT and SURF used as a basis for object recognition techniques used by computer vision systems. The system consists of software embedded in a Raspberry Pi platform, microprocessed by an ARM Cortex A5O que são3 core. O system uses a Linux operating system and has undergone experimenta- tion in the processing time. It has been found experimentally that the processing time by the ORB technique is much faster than the SIFT method. Being that the performance with respect to the rate of correctness in relation to the figure is similar to that of the SIFT SURF. The SUFT method performs in terms of processing slightly lower than the ORB method, but with the great disadvantage of being a algorithm protected by patents.