Processing issues in the automated inspection of power insulators

In this paper, we present an approach to automated inspection of power insulators. In constructing the program, we have analyzed the potential options in processing that affect the inspection process. The location of the insulator is first determined by using a binary template matching process. Then, the disks of the insulator are extracted by a filtering operation. We have compared the performance of three order statistic-based filters for this inspection and found that the median filter performs optimally for our application. Finally, the connected components labelling technique is applied to count the area, width, and height of the disk and the number of disks in the insulator.