Special Issue on Visual Inspection
ProfessorUniversity of Electro-Communications, 1-5-1, Chofugaoka, Chofu, Tokyo 182, Japan
We acquire more than 60 percent of information from our activity environment through our visual sense. The visual sense allows us to collect information about an object from a position away from it without exerting any effects it such as constraining its motion. Visual information acquisition plays a very important role in the industrial field including visual appearance inspection and various other monitoring. A field called machine vision or computer vision has been formed, it is related to the artificial realization and application of the visual function and is now under aggressive study. Inspection using the visual sense, so-called visual inspection, is extremely important; and its automation has been studied for a long time. However, many problems remain to be solved; and in many cases, this operation must rely on human vision. In order to realize the visual function from an engineering point of view, there are many demands for the development of an image sensor that acquires visual information as image information, a method that processes and recognizes image information, and a method that integrates the observation control system allowing processed image information to be systematically organized and the operation to be checked. In consideration of long-term vision as stated above, this special issue provides a description of sensor technology for image information acquisition in the visual inspection process as well as the neural network processing method which is expected as a flexible method for image processing and recognition. For robot sensors, an active method is used to simplify the recognition process, which projects a special light on an object for measurement. This issue includes the topics covering the development of sensors, aiming at their downsizing and high performance. The human visual sense may function by two operating modes: the monitoring mode that senses an unusual situation appearing in the view field and the attention mode that provides detailed analysis of the situation in this area. The former is permitted to have a low detecting, accuracy, but it requires a wide detectable range. The latter is permitted to have a narrow sensing range, but it requires a high sensing accuracy. In other words, multi-resolution sensing operations are performed in the human visual sense. It is desirable for robot sensors to perform the multi-resolution operations that enable coarse sensing to be realized in a wide range and high-accuracy sensing in the attentive area. This issue also includes the development of these sensors. The appearance inspection of welded boats and the recognition of vehicle numbers have been put to practical use, and these topics are also described in this issue. In some cases, techniques visual information processing can make visible to us those that can not be seen by our visual system. This can be thought as an extension of the visual function and the level of sight is very interesting.