The invention relates to a method for identifying a
lightning arrester based on a patrol inspection
robot, wherein the patrol inspection
robot obtains an image of a field
lightning arrester; the target area to be detected is roughly located and precisely located, and the target candidate area is screened to obtain the
lightning arrester image. The digital region is extracted, and the tilt correction and image preprocessing are carried out on the digital region. The target region is segmented to obtain a single
digital image. Image features are extracted and a
template matching method is used to recognize a single
digital image. ehether the matching ratio of a single
digital image and a template is higher than a set threshold value is judged, if so, the recognition is completed, if not, a trained
artificial neural network is called to recognize a single digital image again; the digital display part of the arrester is obtained by sorting the single digits according to their coordinates in the image. The invention can effectively complete the detection and identification task of the
lightning arrester under different illumination and posture conditions, and meanwhile, the identification speed is improved on the basis of maintaining the identification accuracy.