Vol.4 No.2 2011

Research paper−75−Synthesiology - English edition Vol.4 No.2 pp.75-86 (Oct. 2011) 1 IntroductionIn recent years, there have been great expectations for vision systems (computer vision). They are useful in various fields including surveillance cameras for crime prevention, appearance inspection of manufactured goods, CT scans and tissue examination in medicine, robot vision, and analysis and evaluation of movement in sports studies as well as image searching on the Internet. Furthermore, as an important aspect of vision systems, it should be noted that collection and processing of various images has become easier, owing to the technological development of CCD cameras, various sensors, computers, and visualization techniques.With developments in the field of vision systems, image recognition research has been pursued vigorously on an international level, but automation and implementation has been difficult. In addition, only distinct ad hoc methods and expensive specialized systems have been developed, and there is still a reliance on human abilities under actual settings. As a result, the implementation and distribution of a cheap, PC-based vision system that is versatile and delivers high speed is highly desirable.With the above objective, this paper discusses the pattern recognition theory developed by the author thus far[1], focusing on feature extraction theory[2] and the Adaptive Recognition for General-Use System based on it that was proposed as a practical system construction method[3][4], as well as various practical developments[5][6]. Moreover, the effectiveness and importance of the theoretical approach in particular is demonstrated when considering a construction - Its theoretical construction and applications-Nobuyuki OtsuFellow, AIST Tsukuba Central 2, 1-1-1 Umezono, Tsukuba 305-8568, Japan E-mail : Original manuscript received April 14, 2010, Revisions received April 12, 2011, Accepted April 13, 2011In recent years, the need for computer vision systems is increasing in various fields, such as security and visual inspection. It is crucial there to realize simple and high-speed practical vision systems. The present paper addresses the author’s theoretical research and its applications developed thus far in working toward this goal. First, the problem of the conventional approach is pointed out, and the general framework of pattern recognition, in particular the feature extraction theory, is referred to as the theoretical foundation. Next, a scheme of adaptive vision system with learning capability is presented, which comprises two stages of feature extraction, namely, Higher-order Local Auto-Correlation and multivariate data analysis. Several applications are demonstrated, showing the flexible and effective performance of the proposed scheme.ARGUS: Adaptive Recognition for General Use SystemKeywords : Vision system, pattern recognition, feature extraction, adaptive learning[Translation from Synthesiology, Vol.4, No.2, p.70-79 (2011)]method for recognition (generally, information) systems.2 Ordinary approaches and pattern recognitionFirst, let us consider the pattern recognition problem of image measurement and recognition. Fig.1-a) shows the image measurement (enumeration) task where there are round particles of two different sizes and the total number of each is enumerated. The method usually considered is similar to the following sequential method. First, the screen is scanned to segment individual particles, and then the radius is measured for the approximating circle of each particle; in this manner, the size of the particle can be determined from the radius, and the particles can thus be counted. However, this method will clearly result in an increase in the calculation time, proportional to the number of objects.On the other hand, Fig.1-b) is the image recognition problem that identifies what each object (animal) is. It is usual to consider the characterizing features (parts) that distinguish these four objects. In the context of each model, ears, tails, Fig. 1 Examples of vision tasks[6]b) Image recognitiona) Image measurement


page 4