Gravure ink transfer is very sensitive to the presence of small surface depressions in the paper surface. Whenever the area of a depression is close to the diameter of an individual gravure cell, ink fails to transfer, resulting in a white void (missing or skipped dot). For process control or coating development purposes, it is necessary to predict and measure the seventy of these missing dots. Using optical microscopy, human vision has been employed to do this counting, however this is prone to subjectivity and poor precision. Currently available machine vision algorithms usually fail to adapt to different image configurations and filtering anomalies. Therefore, each time an analysis is performed, the algorithm needs to be tuned to the current screen angle, dot size and frequency, which leads to inconsistent rejection of anomalies. We describe a machine vision algorithm that adapts to printing parameters (dot size and frequency) as well as to lighting conditions, to yield an accurate, reproducible result across a wide set of image characteristics, Efficient rejection of anomalies, such as merged dots and partial dots, is achieved through a combination of a conventional mathematical morphology operation known as "dilation" combined with an automated examination of initial dot sizes, and an iterative histogram analysis procedure.
展开▼
