Power Pole Detection Based on Graph Cut

Power pole detection from images is an important problem in the future electric power industry application. A precise detection is essential to inspect the defects of a power pole. In this paper, we propose a novel approach to detect the power pole object from images. Graph cut for image segmentation is a newly developing graph based image segmentation technique. It is effective but takes huge computation burden. The proposed approach combines prior knowledge with graph cut into detection. Firstly, it locates the rough region of the power pole to obtain two restricted regions based on prior rules. Then, a traditional graph cut framework is used in the restricted regions to improve the precision of segmentation. Experimental results verify its efficiency and accuracy.