Machine Vision: Approaches and Limitations

Abstract

Machine vision in its common definition is a possibility of a machine (by sensing means and computer mathematic processing consecutively) to obtain an information about surrounding environment for further analytical treatment. According to this common definition we can unite in a general classification various, sometimes quite different by its principle, technical systems. These classification tables can be represented on the base of two different approaches: 1) practical causes (Soini, 2001) for necessity to “see surrounding environment”, and 2) technical principle and means using for this task solution. According to the common definition any complete Machine vision system combines two components: technical means (or hardware) and information processing mathematics and algorithm (or software). However, the various software analyses is not expedient in view of variety of mathematical methods and their object focused applications in each case (Mordohai & Medioni, 2006); and finally can’t give clearer problem understanding. We are now observing a rapid growth of 3D software and hardware capabilities for mainstream PCs, and 3D graphics accelerator boards with processing capabilities of roughly millions polygons per second are becoming commonplace (Petrov et al., 1998). At the same time, dynamic level-of-detail algorithms—built into standard 3D software packages— offer considerable acceleration of model viewing and progressive loading and transmission of 3D models. Despite the fast growth of computer 3D visualization capabilities, until recently data input technology has remained unchanged. So, in our research for better understanding what is Machine vision, what is its modern state, which practical and technical tasks it decide, and which objective limitations and open problems recently it have, we’ll based on the two mentioned above approaches. In a part of practical reasons, which caused for necessity to develop Machine (or computer) vision concept, can be mentioned: security problems in static/dynamic image analysis in perimeter/volume protection (motion/dangerous object detection); (Chellappa et al., 2005), (Itti & Baldi, 2005) analysis of short/long term deformation of important engineering structures (more commonly known as ‘structural health monitoring’ or SHM); (Athavale et al., 1990), (Allen et al., 2005), (Mallet et al., 2004), (Tyrsa et al., 2004), (Ohno et al., 2002), (Benedetti et al., 2004), (Slob & Hack, 2004), (Stewart & Tsakiri, 2002), (Liwen Dai et al., 2002) O pe n A cc es s D at ab as e w w w .ite ch on lin e. co m