Nowadays, New Zealand families are still using traditional analog water meters. Even if the traditional water meter has the characteristics of accuracy measurement precision, simple structure, zero power dissipation and high reliability, it still needs a worker to record the meter readings door-by-door. With the development of image processing technology, people spend a large amount of research in the field of character recognition, a growing number of character recognition products are applied to industrial production and daily life. In this thesis, an accessory will be designed for water meters by using optical character recognition technology to discern the meter reading. This accessory can translate the analog meter reading into a digital signal, prepare for the automated meter reading system, and Internet of Things in the future. The accessory will sit on top of a conventional water meter, using a camera to capture the number, and display the reading on an additional LCD, which can be sent wirelessly to a database server. This provides a significant alternative meter reading, reducing human resource waste. The contribution of this thesis is a novel light-weight optical character recognition algorithm and its implementation in a prototype with embedded digital system. Experimental results show that the reading accuracy can be as high as 95%, which indicates a promising candidate technology for automated meter reading.