Uma arquitetura para o desenvolvimento de aplicações de visão computacional e processamento digital de imagens em sistemas embutidos

Abstract

This paper describes an architecture of hardware and software that helps the development of applications of Computer Vision and Digital Image Processing in Embedded Systems, specically Smart Cameras. Smart Cameras are cameras that, in addition to image capturing, is capable of extracting application-specic information. Embeddedsystems are systems composed of hardware and software designed to perform specic tasks combining processing and input/output devices needed for applications. Computer Vision can be dened as the ability to rebuild, automatically, the mathematical model from one or more images of a given scene. The mathematical model, or computer graphics, is capable of providing information about the scene previously registered only in images. This is the reverse process to Computer Graphics where the computing system, from a mathematical model, generates digital images. The area of digital image processing is thus the basis for the development of any application of Computer Vision, because it is responsible for the manipulation of images that are the rst input of anapplication of Computer Vision. The restriction of resources inherent in the systems combined with the complexity of Embedded Computer Vision hinders the construction of applications embedded in these two contexts. The solution, proposed in this paper, is to build an architecture that helps the developer, leaving aside details of EmbeddedSystem and Computer Vision algorithms.