| Nowadays,the modern society increasingly relies on electricity,and the stable operation of the power system is particularly important.To ensure the security and stability of the equipment,regular inspection is necessary for substation.Inspection mission in substation is mainly composed of reading recognition of pointer meter and temperature measurement.Many substations become more intelligent and automated,and a large number of inspection robots come into use.This thesis aims at the existing problems of the pointer meter reading recognition method for the substation inspection robot,systematically studies the methods of circle fitting of meter,image registration with positioning error and line fitting for double-pointer meter,and proposes a robust method of reading recognition method of the pointer meter for substation inspection robot.The main contents of this thesis are as follows:(1)Development of a circle fitting method based on least-square method and RANSAC(Random Sample Consensus).This method can stably fit the dial circle of the meter and effectively reduce the noise with the combination of the least square and RANSAC.(2)Development of an image registration method with positioning error for the substation inspection robot.A multi-stage template feature detection and matching method based on SURF(Speeded Up Robust Features)is used to make the images captured by the inspection robot as clear as possible within the allowable range of positioning error.(3)Development of an efficient line fitting method for double-pointer meter.This method is based on the line fitting method for the single-pointer meter,and its validity and robustness are verified through experiments.(4)Development of a correction method of the deformation caused by camera imaging.The pointer images captured by the camera are corrected by perspective transformation to reduce geometric deformation.The displacement error caused by position of the pointer and the dial is analyzed and corrected,reducing the system error. |
PDF Full Text Request