More machine vision applications in the NCEA

Abstract

In the early nineties, the research team at the National Centre for Engineering in Agriculture established a reputation for vision-based automated guidance of agri-cultural vehicles.{4} This work has a new lease of life with recent funding. A succession of further vision projects have been somewhat unusual, ranging from the visual identification of animal species for the culling of feral pigs to vision-based counting of macadamia nuts. A unifying feature is the easy availability of low-cost cameras and a framework for integrating analysis software using DirectX 'filters'. Machine vision has changed from its earlier status as a sophisticated and expensive technology to a low-cost solution for more general instrumentation. For the rapid solution of ad hoc problems, it is easy to exploit the convergence between computing and media entertainment. A system including camera, inter-facing, gigabytes of data storage, display and an embedded PC card can cost less than a single conventional high resolution industrial camera and interface. How-ever we are also pursuing a more fundamental approach, building systems around image-sensor components interfaced by means of RISC processors.