Meter Intelligent Identification Technology Based On Embedded Image Processing

With computer technology and network technology advancing, automated meterreading technology has also achieved an unprecedented development. Automatedmeter reading technology has integrated computer technology, communicationtechnology, network technology and measurement technology, making it convenientto use, accurate to measure and fast to read meter, thus making management easier.Currently, more and more electricity meters, water meters, gas meters etc. haveappeared in people’s daily lives, at the same time, automated meter readingtechnology has also gradually become popular. But that how to make automatedmeter reading technology simpler, more reliable and more efficient has always beenthe hot topic among various researchers and engineers.By analyzing the research status of automated meter reading technology at homeand abroad, and by comparing the advantages and disadvantages of currently usedautomated meter reading technology, this paper proposes a camera reading methodbased on embedded image processing. The target board, this paper designed, takes themicroprocessor S3C2440A as the core, and this paper build ARM-Linux platform bytransplanting embedded Linux operating system onto the target board. Regarding therecognition problems in this meter reading method, this paper proposes an efficientrecognition algorithm, that is, according to the characteristics of meter images, wepreprocess meter images by methods of Gaussian filtering, the maximum betweenclass variance and morphological operations, and then segment dial region and singledigit by methods of contour detection, connected domain segmentation and so on.Finally, identifying the segmented digits based on structural characteristics of digits.The designed algorithms in this paper can effectively recognize the reading onthe meter image, and the built hardware platform has a stable performance. After the application is transplanted onto the hardware platform and tested, we can find that ittakes240ms for the USB camera to take an image and save the bitmap file, and ittakes3130ms to recognize the readings on the saved meter images. This discoverycan meet real-time requirements to a certain extent, and can also lay a foundation forcamera meter reading system.