New detection method of faulty distribution power apparatus using thermal images

In automatic diagnosis of a power distribution apparatus is utilizing thermal images problems arise in that there are problems, that there are many thermal patterns similar to the thermal pattern of the target apparatus and that the temperature around the apparatus influences the diagnosis. In order to solve these problems, we developed a new method whereby images of the apparatus are extracted by an image processing technique based on high-order local autocorrelation features, the attachment pattern on a pole, and a disparity map; a faulty apparatus is identified based on the local temperature gradient. In the extraction method, the distance information provided by the disparity map narrows the search area of the apparatus. The search is conducted according to the rule that an apparatus feature that is defined by high-order local autocorrelation features appears at certain intervals, according to the attachment pattern on a pole, in the search area. The local temperature gradient detects local heat in the form of leakage current on the faulty apparatus. An error rate of 3% was obtained from experiments on the extraction of an apparatus, and an error rate of 17% was obtained from experiments on the detection of a faulty apparatus.