Fault Region Extraction of Electric Equipment Based on PCNN

To localize the fault in infrared images of electric equipment, an efficient fault region extraction method is presented. On the basis of pulse coupled neural network (PCNN), the inner parameters of each neuron is reduced, and local features of fault and non-fault regions are combined to achieve adaptive iteration. Through experimental analyses, this method is proven to be able to extract the fault region more effectively compared with conventional extraction methods based on OTSU, EM, conventional PCNN and normalized cuts.