Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Package Label Detector

IP.com Disclosure Number: IPCOM000118580D
Original Publication Date: 1997-Mar-01
Included in the Prior Art Database: 2005-Apr-01
Document File: 4 page(s) / 92K

Publishing Venue

IBM

Related People

Wolf, S: AUTHOR [+2]

Abstract

Package label detectors typically use a scanner to produce digitized images of packages, often under conditions of changing illuminate color and intensity. In addition to the unknown illuminate and label color, the color of the packages may also vary. A common application is in parcel sorting systems, where the color of the ambient light is uncontrolled.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

Package Label Detector

      Package label detectors typically use a scanner to produce
digitized images of packages, often under conditions of changing
illuminate color and intensity.  In addition to the unknown
illuminate and label color, the color of the packages may also vary.
A common application is in parcel sorting systems, where the color of
the ambient light is uncontrolled.

      Despite the fact that most labels are white, a significant
proportion of colored labels are used.  The effect of unknown and
changing ambient light conditions is to cause even white labels to
appear colored to the image capture system.  This precludes using a
method of finding labels based on absolute color.

      The described method is based on the use of the three color
signals R, G, and B to provide information about color difference.
Fig. 1 shows the R, G and B scan lines derived from a two dimensional
image of a yellow label on a brown parcel.

      As long as there is a color difference between the parcel and
the label, then when photographed through the three "orthogonal"
color filters, there will be a significant intensity difference
between at least one of the color signals and the others.  By
applying a signal dependent threshold to each of these signals, the
label can then be separated from the background.  In the example
shown in Fig. 1, the "R" and "G" signals of the yellow label are more
intense than that of the parcel.

      Fig. 2 shows the resulting planes after applying a global
threshold where pixels above the threshold are classified as class 1
while those below as class 0.  The label can easily be identified.

      After this binarization, the coordinates of the label boundary
can be extracted by contour tracing.

Implementation
  1.  Color Image "Segmentation".  The segmentation algorithm
       consists of two stages:
      o  Color Independent Binarization:  First each of the red,
          green, and blue planes is converted into binary form, in
          which each pixel is classified as either black (class 0)
          or white (class 1).  This process involves the generation
          of a bi-level image from a multilevel one.  In a simple
          example, the greyscal...