Research And Implementation Of Automatic Recognition Of Pointer Meter Readings Based On Machine Vision

In the field of equipment monitoring,pointer meter has strong anti-electromagnetic interference ability and can adapt to high temperature and high pressure and other harsh industrial field environment.It is widely used in the power industry and the industrial manufacturing site with harsh environment.The manual inspection and reading method has high labor cost and low efficiency.When the equipment runs abnormal,it can not be found in time,which is easy to cause safety accidents.With the development of machine vision technology,it has become a new trend to recognize and read pointer instrument automatically by inspection robot.However,the instrument images obtained by the patrol robot have many problems,such as large background interference,various instrument types,and instrument tilt,which make the traditional algorithm unable to adapt to the automatic reading of the instrument by the patrol robot.Aiming at the above problems,this paper studies the dial area positioning,instrument type recognition and instrument automatic reading,proposes the automatic recognition algorithm of pointer instrument reading based on machine vision,and sets up the automatic recognition system of pointer instrument reading.The main research contents include:(1)In order to solve the problem of background interference such as pipeline facilities and instrument buttons in the background of the instrument image,the Mask R-CNN network model in deep learning was introduced to directly detect the dial area.Then according to the shape characteristics of the dial,the anchor frame ratio of the network model is optimized,and the dial partition is designed in the network model.The optimized network model can accurately locate the dial area,and obtain the dial image the corresponding binary mask image without background interference,which lays the foundation for the implementation of the subsequent algorithm.(2)After the dial image is obtained,in order to identify the instrument type and obtain the information of the instrument range,the pointer instrument template library is established,and the instrument type recognition algorithm is proposed based on SURF algorithm.Aiming at the interference of similar feature points in the process of recognition,a method of setting matching point threshold is proposed to eliminate the interference.This algorithm has a high accuracy of meter type recognition,and can provide the information of meter range and scale line coverage for the subsequent automatic meter reading algorithm.(3)Based on the algorithm of dial region location and the algorithm of instrument type recognition,the automatic reading algorithm of instrument is proposed.In order to solve the problem of instrument image tilt during pointer extraction,a perspective transform correction method based on extreme point detection of binary mask image was proposed.Then,the instrument pointer is extracted by the morphological operation of the image.In order to solve the problem that the uneven thickness of the extracted instrument pointer affects the calculation of deflection Angle,the fast parallel thinning algorithm is adopted for optimization.Finally,the Hough transform algorithm is used to get the pointer deflection Angle,and the final instrument representation number is obtained by the Angle method.(4)Based on the algorithm of dial positioning,instrument type recognition and instrument automatic reading,the automatic recognition system of pointer instrument representation number is built,the upper computer interface is developed,and the algorithm of dial region positioning,instrument type recognition and instrument automatic reading are verified by experiments.The experiments show that the algorithm proposed in this paper has high accuracy and real-time performance.