The attachment of foreign objects at key points of the wire can cause short circuits or leakage.To improve the impact of adverse weather conditions and image feature dimensions on visual detection methods for foreign object hidden dangers and reduce false detections,a visual detection method for foreign object hazards in the wire based on adversarial generation networks and key point visual tracking model is proposed.This paper uses the two-dimensional Otsu method to segment and remove fog wire images,extract the target area of wire foreign objects.Based on the segmented images,we use adversarial generation net-works to achieve visual detection of wire foreign object hidden dangers.The experimental results show that the tracking posi-tions of each key point on the wire by using this method are very close to the actual position,with a maximum deviation of only 0.03 m.After defogging,the clarity and color information of key point images on the wire can be significantly improved.It can accurately and completely extract the target area of foreign objects in the wire,and the boundary information processing is bet-ter.
维普
万方数据