Image analysis – Applications – Mail processing
Reexamination Certificate
1999-12-03
2004-04-27
Patel, Jayanti K. (Department: 2625)
Image analysis
Applications
Mail processing
C382S173000, C382S306000, C209S583000, C209S900000
Reexamination Certificate
active
06728391
ABSTRACT:
TECHNICAL FIELD
This invention relates in general to the field of image processing and, more particularly, to a multi-resolution label locator in an automated parcel sorting system.
BACKGROUND OF THE INVENTION
Automated sorting of parcels is becoming very popular because it reduces labor costs while providing fast and reliable parcel delivery services. However, since parcels rarely have the same size and shape, automated parcel sorting that employs image processing to identify address labels becomes very complicated and may be prone to label reading errors.
To capture an image of an address label of a parcel with sufficient quality for a human operator to read and then to key-in the destination address, a camera must scan the surface of a parcel at a relatively high resolution. A high resolution image results in large parcel images and correspondingly large data storage requirements. One problem in the automatic sorting of parcels is processing high resolution parcel images at a rate equivalent to the output of the mechanical section or conveyor system of the automatic parcel sorting system.
In addition to large image processing time, another problem in high resolution image processing of parcels is locating the destination address label. Even with high resolution images, the human operator must still look up, down, or across a screen displaying the image to identify the location of the destination address label. Such eye scans significantly reduce the efficiency of an automatic parcel sorting system.
Other automated parcel sorting systems have attempted to improve efficiency by eliminating the need of a human operator to read and key-in destination addresses of a label. Such other automated parcel sorting systems include devices that employ fiduciary markings and systems that rely on the leading edge of packages having a known shape.
Automated parcel sorting systems that employ fiduciary marks use optical character recognition (OCR) to ascertain the location and orientation of an object or text affixed to an object. For example, an OCR reader system scans a parcel bearing a fiduciary mark and locates the fiduciary mark. In this manner, a fiduciary mark which is placed in a known relation to the destination address block can be used by the OCR system to locate the position of the destination address block. Similarly, an orientation specific fiduciary mark whose orientation is placed in a known relation to the orientation of the text within a destination address block can be used by an OCR system to ascertain the orientation of the text.
While fiduciary mark systems may improve efficiency, these systems require each parcel receiving site to have identical fiduciary markings so that each OCR system can recognize a particular fiduciary mark. Therefore, such systems generally require preprinted labels or parcels comprising the fiduciary mark and specifying a markable area for placing text. Preprinted labels and preprinted parcels are expensive and some percentage of customers will inevitably fail to use them.
For other systems that do not employ fiduciary marks and preprinted labels, the leading edge of parcel with a known shape is utilized to determine the orientation and location of text on a parcel. However, similar to the fiduciary mark systems, these systems do not afford flexibility in the size and/or shape of parcels.
Accordingly, there is a need in the art exists for an automatic parcel sorting system that can readily identify destination address labels within a scanned image of a parcel, regardless of the size and/or shape of the parcel. There is a further need in the art for an automatic parcel sorting system that significantly decreases the amount of time required to process an image or to acquire destination address label data from a scanned image.
SUMMARY OF THE INVENTION
The present invention is a multi-resolution label locator that provides a list of one or more areas within a processed image of a parcel that may contain labels of interest. The multi-resolution label locator is typically part of an automatic parcel sorting system.
The automatic parcel sorting system typically includes a video camera mounted adjacent to a conveyor apparatus. The video camera is operatively linked to two video processors, which produce at least two different kinds of image signals of a parcel. The video processors produce a first decimated (low-resolution) image of the parcel and a second image that corresponds to edge-occurrences of indicia expected to appear on a label, such as text.
The two images produced by the video processor identify different characteristics of the original high resolution image. For example, the decimated-image hardware of the video processor may identify areas in the image that have characteristics typical of labels, whereas the edge-occurrence processor may identify areas that have characteristics typical of text.
The two images are fed into a separate microprocessor, which employs a multi-resolution label locator program to identify one or more areas on the parcel that may contain a label of interest. The multi-resolution label locator program then classifies these areas and compiles a list of these candidate areas based on data extracted from the first and second images produced by the video processor.
Generally stated, the invention is a multi-resolution label locator system for an automatic parcel sorting system. The multi-resolution label locator system obtains a video signal containing a plurality of pixels that define an input image of a substrate. The multi-resolution label locator system divides the input image into a plurality of multi-pixel cells. In subsequent computations, the multi-resolution label locator system extracts feature values corresponding to the preprocessed decimated image and edge-occurrence image.
The multi-resolution label locator system then creates the decimated image (low resolution image) corresponding to the input image in order to reduce the amount of data in the subsequent computations. This decimated image is generated by utilizing a common-characteristic value, such as a single pixel, that corresponds to each multi-pixel cell of the input image. Each common-characteristic value represents a decimated image of the pixels within the corresponding cell. For example, if the multi-resolution locator system is designed to locate labels on a package or parcel, then the system will look for large, relatively white contiguous areas (or areas having a different color depending on the operating environment of the present invention) on the package or parcel since labels generally have a different color or reflect light at a different intensity relative to the package or parcel. Those regions of the parcel or package having a higher light intensity or different color value are assigned a decimated-image value and this data is then mapped to an image space to create the decimated image.
With this decimated image, the feature extraction function implemented on the microprocessor can efficiently extract feature parameters of the label candidate areas. Some of the feature parameters may include: normalized dimensions and areas of the label candidates, aspect ratios, and the relative average light intensities of potential label candidate areas derived from the decimated image. These feature parameters become the input data for the classification function (also discussed infra).
While the first video processor of the multi-resolution locator system is generating the decimated image, the second video processor of the multi-resolution label locator system simultaneously creates an edge-occurrence image that corresponds to the input image. The edge-occurrence image includes an edge value that corresponds to each cell of the input image. Each edge value represents the number of occurrences of edges within the pixels of a corresponding cell of the input image. For example, if the multi-resolution locator system is designed to locate address labels on a package or parcel, the locator system will look for closely spaced black and white transit
Recktenwalt James
Skinger Gregory P.
Wang Jason
Wu Jianxin
Alston & Bird LLP
Carter Aaron
Patel Jayanti K.
United Parcel Service of America Inc.
LandOfFree
Multi-resolution label locator does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Multi-resolution label locator, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Multi-resolution label locator will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3215757