Techniques and applications of electrical equipment image processing based on improved MLP network using BP algorithm

This paper proposed a new method combining infrared and visible image, which is devoted to monitor the electrical equipment. The key technologies of infrared and visible image processing include two aspects. On the one hand, it researches on targeting localization based on visible image. We can get the horizontal and vertical adjustments of the refitted cradle head camera through the coordinates of the electrical equipment in the image by means of the strong ability for BP algorithm fitting nonlinear function; on the other hand, the infrared image is processed by qualitative method based on the temperature difference The improved MLP network eliminated the harmful effect of the size of the partial derivative on the weight step. The network received some features as the inputs and classified the conditions of equipment into two categories which are defective and normal. Experimental results show that the positioning error is less than 5 pixels and the classification accuracy rate reaches 84.21%, which means that the method not only improves the measurement efficiency of the infrared detection technology, which is not affected by the limitation of measurement environment.