We need to be able to recognise numbers in an image containing multiple numbers (or one number and parts of other numbers).
We therefore need an algorithm which returns an image containing only the text (i.e. number and spot) for one number.
This could be done by
finding contours
grouping contours by proximity (e.g. by bounding box centre, or shortest distance between contours)
To begin with, we can probably assume that the desired spot is fairly central in the image. So if there are multiple spots, just consider the most central. A further enhancement could try to determine whether a spot in the image also has it's number fully in the image and so recognise multiple spots in the image.
JackBuck
commented
7 years ago
We need to be able to recognise numbers in an image containing multiple numbers (or one number and parts of other numbers).
We therefore need an algorithm which returns an image containing only the text (i.e. number and spot) for one number.
This could be done by
To begin with, we can probably assume that the desired spot is fairly central in the image. So if there are multiple spots, just consider the most central. A further enhancement could try to determine whether a spot in the image also has it's number fully in the image and so recognise multiple spots in the image.