A Recognition Method of Action-Times-Readout of Arrester in Convertor Station Based on Image Processing

摘要

According to the characteristics of the converter in the power system for lightning arrester action frequency instrument, presents a recognition method for lightning arrester action times based on image station. First of all, using perspective transform method for image distortion correction, secondly, to enhance the image by image grayscale, image smoothing, histogram equalization, image of two value method and image expansion and refinement, the image information and character relatively clear the position, again, using vertical projection method and neural network based on sport to conduct image segmentation and character recognition, and finish the reading of the lightning arrester action number recognition. Finally, the scene of the image of the proposed algorithm are verified. The results show that the algorithm can identify fast, automatic, accurate number of arrester. This method can avoid the artificial reading efficiency is low and the problem of poor accuracy, but only the intelligent recognition of very few changes can be applied to other digital instrument in converter station.