Optical Character Recognition for Intelligent Meter Reading

OData support
Supervisor:
Dr. Szűcs Gábor
Department of Telecommunications and Media Informatics

One of the most significant discipline of the past few years is Artifical Intelligence, which can help us to replace the manually created functions into automated, or partially automated solutions.

OCR (Optical Character recognition) is one of the most significant achievements in this area of science, which enables recognition of optical characters from images. One of the most famous OCR engine called Tesseract, which reached success in reading contigous characters.

In my thesis I developed an images processing system, which is able to detect and crop the digits of the gasmeter’s value from images are being taken of gasmeters and classify these cropped digits with a classifier based on neural networks. In order to do so, I created an algorithm that can automatically detect the digits tied to the gasmeter’s reading from the surrounding rectangles. Besides this it can recognize the serial number and the barcode’s string of the gasmeter.

The digit classifier, I used to build this model, was taught by the MNIST dataset, which includes labeled digits similar to the digits of gasmeter’s readings. The testing was performed on individually collected gasmeters in several phases per subtasks.

Downloads

Please sign in to download the files of this thesis.