CPC G06K 7/1417 (2013.01) | 16 Claims |
1. A computer implemented method for detecting and reading a barcode in an image of a marked glass substrate, said method taking as input a raster image of at least one portion of the marked glass substrate and providing as output an abstract image of said barcode, and said method comprising the following steps:
(a) computing a probability occurrence θ of said barcode in the raster image with a first trained convolutional artificial neural network;
(b) computing coordinates of representative points of a boundary of the barcode in the raster image according to a given threshold σ for the probability occurrence θ computed at step (a);
(c) cropping the raster image to the boundary of the barcode according to said coordinates computed at step (b), and
(d) computing with a second trained convolutional artificial neural network a tensor of probabilities for each pixel of the cropped raster image, said tensor being the abstract image of the barcode.
|