Reading A ConEd Meter From An Image

Takes an image of a Con Edison dial meter and outputs the number reading to give to your Con Edison representative.

Requirements

Python 2.7.10
NumPy 1.10.1
OpenCV 2.4.12

Setup

Create and activate a new virtualenv:

$ virtualenv mr-env
$ source mr-env/bin/activate

Install requirements:

$ pip install -r requirements.txt

Install OpenCV using Homebrew.

$ brew install opencv

Navigate to mr-env/lib/python2.7/site-packages and create a symlink to your version of OpenCV's Python modules. The path may change depending on the location of your Open CV python packages.

$ ln -s /usr/local/Cellar/opencv/2.4.12/lib/python2.7/site-packages/cv.py cv.py
$ ln -s /usr/local/Cellar/opencv/2.4.12/lib/python2.7/site-packages/cv2.so cv2.so

Usage

First run your virtualenv, then run the program:

$ source mr-env/bin/activate
$ python read_meter.py -i sample-data/meter2.jpg -d 5

Use flag -i or --image to indicate the path of the image. In this case, the path of the image is sample-data/meter2.jpeg.

Use flag -d or --dials to indicate the number of dials you are looking for in the image. In this case, the number of dials is 5. If no flag is given, this defaults to 4.

Image specifications and limitations

For accuracy in readings, images must be:

scalable to either 480x640 or 640x480 without significant distortion
100% of the meter must be visible
meter should be photographed head on and not from an angle, so that the dials still appear as circles in a straight horizontal row parallel to the top and bottom edges of the photograph.

Examples of good images can be found in the folder sample-data. Examples of bad images can be found sample-data/unsupported.

Currently only supports meters where dials are in a straight row, and not in an arc.

TODO: