Research On Instrument Reading Recognition Based On Substation Inspection Robot

With the increasing dependence of society on electricity in production and life,how to effectively ensure the production safety of power systems,especially substations,has become a hot research topic.The inspection of substations is mainly the inspection of various types of instruments.Inspection is mainly done manually.These methods have low efficiency and accuracy,and cannot adapt to the increasing number of monitoring equipment such as meters and increasingly complex scenarios of substation environments.Therefore,the meter reading recognition scheme based on inspection robot with higher degree of automation,higher accuracy and efficiency has very important research significance.Aiming at the problems that the existing meter reading recognition schemes cannot meet the real-time requirements and cannot cope well with the recognition of various types of meter readings in complex environments,this paper has designed and designed meter reading recognition schemes based on deep learning technology.The main research contents include:1)Aiming at the situation where the instrument is in a more complex environment,an improved instrument dial detection method based on YOLOV3 network is proposed.The traditional computer vision scheme and deep learning scheme in the target detection method are researched and analyzed,and the YOLOV3 network based on the Mobile Net V1 network improvement is designed as a representative of the deep learning scheme.The two schemes are compared to verify the designed deep learning.Advantages of the program.2)Aiming at the rich information contained in the dial,an instrument dial information extraction algorithm is proposed,which mainly includes an image preprocessing algorithm,a YOLOV3 network improved from the Mobile Net V3 network and Trident Net network,and an indicator pointer area extraction algorithm.The YOLOV3 network is used to extract range and dimensional information.3)The reading recognition algorithm of pointer and digital display instruments is proposed.The dial indicator precise positioning algorithm and reading recognition algorithm are designed for the pointer instrument research.The dial fine positioning algorithm uses the center line of the pointer area to extract the algorithm and tilt correction.The algorithm completes the precise positioning of the dial information,the reading recognition algorithm combines the scale distance method and the angle method to complete the reading recognition,and the logarithmic display instrument directly completes the reading through the combination of the range number and the dimension.4)The overall plan of the meter reading recognition and the user management interfaceare designed,and the overall plan is tested systematically to verify the feasibility of the overall plan.