摘要: It is well understood that artificial vision enables a wide range of applications from visual inspection, measurement, recognition, surveillance, to guidance robot systems in real-time and real environment. However, the literature, there no definite answer what (or robot) should be, or how it be. This dilemma largely due fact being actively pursued by scientists various backgrounds both social natural sciences. In this paper, intention present new way re-organizing concepts, principles algorithms vision. particular, we propose function-centric view comprising these five coherent categorizations, namely: (a) instrumental vision, (b) behavior-based (c) reconstructive (d) model-based (e) cognitive abandons long-standing notions low-, intermediate- high-level as they are more illusive than insightful. The contribution article two-fold. First, time assess consolidate current achievements Secondly, important objectively state remaining challenges order guide future investigations