Automatic Number Extraction from Fixed Imaging Distance


Developed countries are facing many challenges to convert large areas of existingservices to electronic modes, reflecting the current nature of workflow and theequipment utilized for achieving such services. For instance, electricity billcollection still tend to be based on traditional approaches (paper-based and relyingon human interaction) making them comparatively time-consuming and prone tohuman error.This research aims to recognize numbers in mechanical electricity meters andconvert them to digital figures utilizing Optical Character Recognition (OCR) inMatlab. The research utilized the location of red region in color electricity metersimage to determine the crop region that contain the meters numbers, then extractsthis numbers region and convert it into binary image and extract the numbers as atext using OCR technique.A camera for the Iphone 6 (8-megapixel) is used to take a snapshot of the meterscreen. The red box in the meter is used to calculate the window coordinates(vertical and horizontal length) that contain the numbers in the original image. Theresults show a high level of accuracy, reaching 100% due to the effort done on preprocessingthe digital images before feeding the part that contains the numbers intothe OCR engine. Compared to the maximum accuracy obtained in other previousresearch of less than 100% in most of related works, the suggested method providebetter approach to obtain the optimum results .Despite the strong results, some challenges still need to be investigated further tofind the best solutions, for example the issue of scratched or unclear meter screen,also the meter type 2 (the type that do not have the red box)