Insulator inclination positioning and identification method based on R-DFPN algorithm

Abstract

The invention discloses an insulator inclination positioning and identification method based on an R-DFPN algorithm. The method comprises the following steps: firstly, acquiring insulator data by adopting an aerial photography unmanned aerial vehicle, and performing data enhancement and manual labeling to obtain a data set; meanwhile, wherein in the training process, a dense feature pyramid is used for obtaining multi-scale feature information, and acquiring a target box with an inclination angle through a rotation candidate region extraction sub-network; outputting classification and positioning information through a classification regression sub-network, and finally outputting final classification information and positioning information with the inclination angle through inclination non-maximum suppression so as to achieve the purpose of detecting five pieces of position information and category information of an insulator in an image. The specific position of the insulator is accurately positioned in the picture, irrelevant background information is eliminated, and the method has very important application value in power equipment inspection.

Description

technical field [0001] The invention belongs to the field of computer vision, and in particular relates to an insulator tilt positioning and recognition method based on the R-DFPN algorithm, which realizes classification and positioning of insulators in pictures by means of deep learning. Background technique [0002] In the transmission line in the power system, the insulator is one of the very important components. First of all, insulators laid in the outdoor environment are more susceptible to failures such as wear and bending due to natural factors such as climate interference, coupled with the influence of the material of the insulator itself, after a certain period of operation, some insulators will fail. Furthermore, due to the long-term operation of the insulator, it will also be damaged due to some changes in the internal load relationship. Secondly, most insulators operate in an environment with relatively large temperature differences, because changes in temperat...

AI Technical Summary

Problems solved by technology

The current insulator positioning technology is mainly divided into two categories, one is to use traditional image processing methods, this type of method needs to manually extract adapted features according to specific scenes, and the generalization ability is not strong enough; the other is to use depth The learning method, the deep learning method is more robust, and does not require manual feature extraction

The current deep learning methods are all based on the horizontal axis symmetric bounding box to locate and identify insulators. Since the characteristics of insulators are of high aspect ratio, the shape of insulators in most aerial pictures is inclined, which will cause the detected insulators to contain A lot of background information, causing certain difficulties for the insulator fault detection to be carried out in the next step

Images