earticle

영어

This paper first analyzed the structure and the edge symmetry of insulator and divided the insulator into 8 kinds of edge according to these characteristics. Then it improved the algorithms that segment and label the region of interest (ROI) based on the depressed region of insulator, which makes labeling multi-insulators in a single image solved. To accurately extract the edge of insulator from the ROI, this paper proposed the double thresholds edge detecting method on account of parallax principle in binocular vision, avoiding of the deficiency of under-filtering and over-filtering within the single threshold method based on monocular vision. It analyzed and removed local and global irrelevant pixels using statistical methods, thereby stereoscopically matched the pixels in each ROI and calculated their 3D coordinates. Then 2 vectors are constructed based on the maximum and minimum of the Z coordinate component: the vertical vector and the direction vector of the insulator. Finally, every swing angle can be calculated using the inner product method. Experiment results show that the relative error of this method is less than 1.3%, it not only has higher accuracy than the traditional method, but also avoids the low efficiency of a single swing angle measurement of traditional method.