Method for Recognizing Insulator from Airborne Image

In the paper, for airborne image by helicopter on power grid, we propose a method to recognize tempered glass insulator based on intensity information. This method employs the morphology, connected components extraction and threshold value of mean segmentation. The feature of tempered glass insulator on image is reflected its intensity information. Firstly, the intensity image is preprocessed by reducing noise. Secondly, morphological erosion is applied. Thirdly, we can extract the area by 1 in the preprocessed image as connected components, which include the part of insulator and noise. Finally, by setting the mean value as the threshold value, the connected component is classified into two parts: insulator or not. Experiments illustrated that tempered glass insulator in image can be recognized by the proposed method with little mistakes.