期刊名称：International Journal of Image Processing (IJIP)
电子版ISSN：1985-2304
出版年度：2011
卷号：5
期号：2
页码：145-165
出版社：Computer Science Journals
摘要：Thermal monitoring is useful for revealing some serious electrical problems in a factory that often go undetected until a serious breakdown occurs. In factories, there are various types of functioning machines to be monitored. When there is any malfunctioning of a machine, extra heat will be generated which can be picked up by thermal camera for image processing and identification purpose. In this paper, a new and effective omnidirectional machine condition monitoring system applying log-polar mapper, quaternion based thermal image correlator and max-product fuzzy neural network classifier is proposed for monitoring machine condition in an omnidirectional view. With this monitoring system, it is convenient to detect and monitor the conditions of (overheat or not) of more than one machines in an omnidirectional view captured by using a single thermal camera. Log-polar mapping technique is used to unwarp omnidirectional thermal image into panoramic form. Two classification characteristics namely: peak to sidelobe ratio (PSR) and real to complex ratio of the discrete quaternion correlation output (p-value) are applied in the proposed machine condition monitoring system. Large PSR and p-value observe in a good match among correlation of the input thermal image with a particular reference image, while small PSR and p-value observe in a bad/not match among correlation of the input thermal image with a particular reference image. Simulation results also show that the proposed system is an efficient omnidirectional machine monitoring system with accuracy more than 97%.
