Research On Machine Vision Recognition Algorithm Of The Image Of The Indication Of The Camera Meter

In the process of industrial automation and intelligent community construction,automatic remote meter reading technology has gradually become a key and popular.The display image recognition system in this thesis is a system that uses image acquisition module to obtain the display image of dial wheel,and then uses the positioning recognition algorithm to realize the automatic recognition of the display value.In this thesis,the water meter is selected as the specific object of study,and the research conclusions and results are also applicable to the automatic recognition of other types of mechanical meter indication image.Firstly,the thesis analyzes the advantages of the camera water meter and thedevelopment and current situation of character recognition,and constructs the framework of the water meter image recognition system.In the part of image acquisition,the design of water meter camera module and concentrator communication module is carried out to realize the acquisition and remote transmission of digital image,so as to realize the recognition and output of digital image on the upper computer.Among them,the realization of the recognition algorithm of water meter reading wheel is divided into two parts:positioning and recognition.In the part of automatic positioning algorithm,Gauss filter,Canny edge detection,Hough transform,affine transform,binary processing and projection method are used to realize the positioning and cutting of given water meter digital characters;in the part of character recognition algorithm,a method based on the combination of grid feature matching and template matching is proposed to realize the recognition of whole characters and half characters.Finally,the design and implementation of font template of font library are completed by software;the dynamic link library is created by positioning recognition algorithm,which is convenient to call;the interface of identification result display and user interaction is created and designed by MFC of visual studio 2017,which realizes the display and interaction of parameters and image results.In the part of automatic recognition algorithm of reading wheel,the paper mainly improves and innovates the positioning part and character recognition part of the wheel.(1)The improved P-tile method is used to segment the image.In order to ensure the accuracy of projection positioning,the separation of target and background is realized when background interference is removed as much as possible.(2)The combination of grid feature matching based on binary image and template matching algorithm is used to interpret the indication reading.The realization can not only ensure the correct interpretation of the whole character,but also improve the recognition rate of the half character.At the same time,it can greatly overcome the influence of the spot and other interferences on the result in the character recognition.(3)When the final result of character combination is output,the reading may increase sharply,and the correction processing shall be carried out.Eliminate the influence of the dislocation of the character wheel on the output result.Through the experimental verification,the recognition method designed in this thesis has a high recognition effect on the visual recognizable non serious interference.