TAF PUBLISHING

Automatic water meter reading recognition is the key feature of this paper from different environmental conditions. To do so, we have proposed a real-time automatic water meter reading recognition model in this research paper. Initially, from the meter image, the largest circular blob is extracted by creating a mask. As the water meter reading region can be tilted at a certain angle, the Hough transform is used to detect and correct the tilted angle. After tilt correction, to find out the circular meter region, the image masking technique is used following the largest blob finding technique. The corrected RGB meter image is converted to a grayscale image to reduce the computational cost for further processing. To improve the quality of the grayscale image, global histogram equalization is used. Using the Otsu threshold method, a gray image is converted to the binary image automatically. We have overcome the challenges of clipping the region of interest (ROI) extraction region. Applying horizontal and vertical projection, the region of interest (reading region) is extracted. At this point, a morphological operation has been implemented to reduce noise. To segment the digit image, a vertical projection algorithm is used. These segmented digits are then matched with template digit images following the correlation coefficient values. This work will overcome the problem of traditional water billing procedures and will be helpful to run a central water billing management.