| Substation plays an important role in transforming current and voltage and distributing electric energy in power system;the normal operation of substation is related to the development of national economy and the stability of normal social life order.In order to ensure the stable operation of the substation,it is necessary to regularly patrol and collect data in the station regularly,and discover hidden dangers of the equipment at the first time to prevent problems before they occur.At present,the traditional inspection is based on manual inspection.The efficiency of manual inspection is low,the quality of inspection is unstable,and it is easy to be affected by personal emotions and status.With the rapid development of the disciplines of artificial intelligence and robotics,the emergence of inspection robots can replace inefficient manual inspections and lay the foundation for unattended substation models.This paper takes the substation inspection robot as the background,and carries out systematic research on robot positioning technology,instrument panel detection,image restoration,image correction,and intelligent meter reading.The main research contents and innovations of this paper are as follows:(1)Introduce the indoor and outdoor positioning technology of the inspection robot;among them,the analysis focuses on the visual SLAM framework.In this paper,the visual SLAM technology is applied to the indoor inspection.In view of the limitation of the visual SLAM positioning,the AR-Tag-based visual positioning algorithm is proposed;Set up an experimental platform and verify through experiments that the algorithm in this paper can effectively correct the positioning error of visual SLAM;(2)Set up a deep learning environment,use CenterNet network to detect and locate the dial and pointer position.The test results show that the dial detection accuracy rate reaches 100%,and the pointer detection accuracy rate is 98.2%;Aiming at the disturbance of the mirror surface of the instrument panel,the self-made data set is used to train the CycleGAN network to visually evaluate the repair results.The method in this paper can effectively remove the obstruction of the mirror surface of the dial,and the brightness and contrast of the image are also improved;(3)Due to the angle of shooting and other reasons,the dial is deformed.This paper designs an image correction algorithm based on feature points.Among them,experiments and analysis are performed on SIFT,SURF,ORB feature point detection algorithms,and finally SIFT feature point algorithm is used to achieve image correction.Experiment and evaluate the dial correction algorithm.The experimental results show that the image correction algorithm in this paper can effectively correct the image deformation and lay a good foundation for subsequent automatic meter reading;(4)A linear fitting algorithm based on feature projection is proposed to extract the pointer.In this paper,the experimental comparison between the least squares straight line fitting algorithm and our algorithm is performed.The experimental results show that the proposed algorithm has a significant effect on pointer fitting,can effectively resist noise interference,light and shadow interference,and has high robustness.;(5)A circle centering algorithm based on MeanShift clustering is proposed to detect the center position of the dial.Finally,based on the tick marks,the center of the circle,the pointer and some prior information,the angle method is used to implement automatic meter reading technology.The different dial samples are selected for experimental verification.The accuracy of automatic meter reading in this paper is high,and the final error is controlled within 3.0%,which meets the accuracy requirements of meter reading in substations. |
PDF Full Text Request