Research On Pointer Instrument Recognition Algorithms For Orbital Patrol Robot In Electric Power Room

In order to build an intelligent and unmanned operation and management machine room,patrol robot is gradually used to replace manual work in the electric power room.But in the process of patrol inspection,the automatic recognition of pointer instrument always has the problems of low recognition accuracy and susceptibility to the influence of illumination changes.In this paper,the problem of pointer instrument recognition is studied.The methods of instrument detection and location,pointer and scale fitting and instrument scale fitting in complex environment are systematically studied.A pointer instrument reading method suitable for electric power room inspection robot is realized.The main work and innovation of this thesis are as follows:(1)Instrument detection algorithm based on Faster RCNN and similarity measure.Due to the inaccurate selection of candidate positions generated by Faster RCNN network,it is easy to cause problems of false detection and missed detection.The template matching strategy based on feature similarity comparison is used to analyze candidate regions and classify them from the feature level,which further strengthens the selection mechanism of candidate regions,helps to reduce false detection and missed detection,and improves the detection accuracy.(2)The algorithm of reading recognition for pointer instrument is studied.The tilt angle of the instrument is calculated by edge detection algorithm based on contours to complete the calibration of the instrument.The gradient second derivative Canny edge detection algorithm is used to filter the non-scale line segment,LSD algorithm is used to identify the scale line segment,then Zhang fast thinning algorithm is used to refine the scale and skeleton branch points are used to extract the center of gravity of the scale line.Finally,the relationship between scale and pointer is analyzed by angle method and distance method,and the indicator number of pointer instrument is calculated.Finally,the results are analyzed through experiments.Both methods can ensure low error rate,meet the requirements of inspection tasks,and improve the efficiency of reading recognition of pointer instruments.(3)The instrument reading recognition algorithm in complex environment is studied.Firstly,an adaptive Gamma enhancement algorithm based on Retinex theory is used to deal with the influence of uneven illumination on image surface.Then,after the scale is extracted,there exists the default of the scale.According to the angle relationship between the scales,the interpolation method is used to repair the default.(4)The design of automatic identification system for pointer instrument is studied.First,the overall design of the inspection robot system is described in detail.Then,the hardware and software of the system are introduced,including the user graphical interface,each step of the system algorithm flow,and the visualization is carried out.Finally,through experiments and results,the recognition results of the algorithm are analyzed,and the real-time performance and effectiveness of the algorithm are verified.