In the 3 previous articles I explained how to develop a really simple barcode decoding algorithm. This program was really naive and only worked with really clean barcodes.
As an assignment for a school project, a more advanced version was developed. The program can now detect and decode EAN13 barcodes in large images with various objects and different backgrounds.
Demonstration
Here are the output images of the program. Red rectangles indicates that the segmentation phase detected a potential barcode but the decoding phase rejected it. Green rectangles indicates that a barcode was found and correctly decoded (using the check digit equation, we can detect most decoding errors).
Click on the images for a better view of the output.
As you can see, there were some vicious images :), especially the one with the wooden floor that totally messed up our segmentation, fortunately, the decoding process rejected all those potential barcodes.
The barcodes themselves are sometimes really noisy. If you zoom on some of them you will notice there are large changes in illumination, a low resolution, low contrasts... These problems required an approach much more robust than the one described in the previous articles.
Comments are closed.