Support Vector Machine based automatic electric meter reading system

The traditional method of manual electric meter reading is very tedious and is prone to lot of errors and has a lot of disadvantages. Some of these disadvantages include low efficiency, man power consuming. The existing methods of Automatic Meter Reading are based on measuring the electric impulse of the sensor. This is prone to wrong counting of the impulses which leads to faulty meter reading. And so a better option is to fit a image acquisition device like camera in front of the meter that will take realtime pictures of the meter readings. This picture is then processed, segmented and the individual digits are recognized using unsupervised feature learning technique-Support Vector Machine. The advantages of this method is that it can generalize over the large degree of variation between styles and recognition rules can be constructed by example. This highly efficient classifier is used for both detection and recognition of these digits.