Research And Application Of Pointer Meter Reading Recognition Algorithm Based On Key Point Detection

The electric power industry is the foundation to support the sustainable and rapid development of society and economy.In order to promote the power industry to better meet the needs of economic and social development,it is necessary to improve the efficiency of power construction,production and consumption.As the core hub of electricity from the production side to the consumption side,improving the daily operation efficiency of the substation has great significance for the rapid development of the power industry.Among them,realizing automatic reading recognition of a large number of pointer instruments in a substation and replacing manual meter reading is one of the keys to improve the operating efficiency of the substation.Based on the summary of key point detection and traditional pointer meter automatic reading recognition methods,this paper designs a deep learning-based key point detection algorithm and applies it to pointer meter automatic reading recognition tasks.The main research results obtained in this paper are as follows:1)An accurate key point detection algorithm based on Mask RCNN is proposed.By adding bias branches to the Mask RCNN network structure,the key point position deviation caused by upsampling in the original network is corrected,and the positioning accuracy of the network on key point detection tasks is improved.Experimental results on public data sets show that the algorithm proposed in this paper has improved on various indicators,effectively improving the accuracy of key point positioning.2)An automatic reading recognition algorithm of pointer instrument based on key point detection was proposed.This paper creatively combines the traditional pointer meter automatic reading recognition task with deep learning-based key point detection tasks,and designs a pointer meter automatic reading recognition algorithm framework based on accurate key point detection.The key point detection algorithm is used to locate the tick marks and pointer position for further reading recognition.The experimental results show that the algorithm proposed in this paper has high accuracy for reading recognition of various pointer instruments,and is very robust to changes in shooting angle and lighting conditions.