Image analysis – Image transformation or preprocessing – Changing the image coordinates
Reexamination Certificate
2000-06-02
2004-04-06
Wu, Jingge (Department: 2623)
Image analysis
Image transformation or preprocessing
Changing the image coordinates
C382S141000, C382S173000, C382S300000, C382S282000, C358S451000, C358S525000, C348S240200, C348S561000, C345S660000, C345S667000
Reexamination Certificate
active
06718074
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of Invention
This invention relates to inspection systems and methods for machine vision applications, and more particularly relates to techniques and systems that provide improved processing of under-resolved data for object feature detection.
2. Description of Related Art
Machine vision systems are increasingly employed to augment or replace human vision in a wide range of industrial manufacturing processes. Machine vision inspection systems are presently being used in such industries as biomedical, automotive, semiconductor wafer and semiconductor chip inspection. A machine vision system typically provides computer-based image processing capabilities that may or may not be automated and can be customized for various vision tasks, for example, machine guidance, part identification, gauging, alignment, or inspection tasks. In such tasks, a machine vision system captures an image of a sample-object of interest and processes that image to produce sample-object information such as position, size and orientation.
Machine-vision systems and methods typically use an analog or digital video still-frame camera to take a picture(s) of a sample-object and provide greyscale or color object image-data corresponding to features of the sample-object. This object image-data is typically provided to a computer display for review by an operator, or output to equipment handling the object. For example, object image-data may be sent to bells, lights or other devices that may alert an operator to a problem. Alternatively, the object image-data may be processed resulting in directives sent to a mechanical diverter that removes defective parts from a manufacturing line, or to a positioning table that moves sample-objects into position prior to a manufacturing step. Additionally, the object image-data may, for example, be processed for comparison against a known model or template to determine the quality of a sample-object under inspection.
An important image processing capability typically provided by machine vision systems is object and feature edge detection for automated inspection. Feature edge detection involves processing greyscale object image-data to extract the edge features of the object under inspection. Object feature edge detection may generally be achieved by extracting selected information from an object image and evaluating the selected information to locate edgelets, i.e., edge points or edge vectors within the object image-data. Edgelet information, e.g., edge point magnitude or edge point angle, can then be used as the basis for determining, object feature size, position or identification. For example, see “Quantitative methods of edge detection,” by I. E. Abdou, in USCIPI Report 830, Image Processing Institute, Univ. Southern California, July 1978.
One of the fundamental challenges for machine vision object and feature edge detection is the robust and accurate edge detection of under-resolved object image-data. When a sample-object under inspection comprises a minimal number of features that are of a relatively large size in comparison to the image field of view, the resolution of the object image-data may be set such that a requisite resolution is provided so that the features may be effectively inspected. However, when a sample-object under inspection comprises many features in a single field of view that is significantly larger than the size of the features, conventional feature edge extraction methods and systems can fail to provide object image-data resolution sufficient for reliable and accurate feature edge detection. As a result, although conventional feature edge extraction methods and systems may provide quality object image-data when operating within their performance bounds, the available image-data resolution is sometimes insufficient to provide for accurate edge detection and subsequent feature-based inspection.
Conventionally, if an object or feature that is less than seven or eight pixels along any axis, the image pixel data including the object or feature is generally classified as being under-resolved. Conventionally, such under-resolved data is problematic in performing boundary-inspection techniques because quality feature edge data cannot be reliably obtained from features that are insufficiently represented in the object image-data.
Although various conventional techniques have been used to compensate for under-resolved data, these techniques are time-consuming and can introduce unacceptable cost to performance tradeoffs. One such technique requires installing higher resolution cameras as part of the inspection system. However, such installation can be extremely costly for large manufacturing facilities with multiple inspection lines. Another technique involves manipulating the physical optical elements used in the camera that obtains the object image-data, e.g., moving a lens closer to the object to provide better resolution of smaller features under inspection. Such a technique is commonly understood to be a physical optic zoom.
However, a physical optic zoom requires, at the least, a sufficient amount of time to adjust the physical optics elements during an inspection process. Such an adjustment may also require a more complicated camera setup, i.e., one having the capability to perform pre-programmed physical optic zoom using a motorized depth and lens control. A faster non-mechanical alternative to physical optical zoom is to use multiple fixed optical setups that provide redundant possibly overlapping views of different portions the field of view at varying magnification factors. This approach eliminates the runtime motor adjustment at the cost of multiple conventional cameras. Depending on the physical constraints of the inspection process this approach may provide a valid alternative. Usage of digital zoom provides an alternative to these physical methods that leverages the steadily increasing processing capabilities of modem computers. This is in contrast to the physical options described that have remained relatively fixed in cost over that last decade.
SUMMARY OF THE INVENTION
The exemplary embodiment of the invention provides a system and method that increase the resolution of object image-data without altering the physical optics of cameras used during inspection. The exemplary embodiment of the invention provides a system and method that perform what may be qualitatively characterized as a software or digital zoom effect. Using this technique, the resolution of object image-data provided by a camera may be reliably increased by a factor ranging from 100 to 400%.
Moreover, in accordance with the exemplary embodiment of the invention, a determination may be made regarding whether portions of the object image-data are under-resolved. Then based on that determination, a local digital zoom may be performed on that under-resolved portion or sub-window of object image-data to increase the resolution of that portion of object image-data. Additionally, the resulting portion of the object image-data with artificially increased image resolution may be scale normalized. This step allows the portion of increased resolution object image-data to be combined with the other non-zoomed object image-data to provide an accurate representation of the original sample-object and/or sample-object at the original working image resolution.
The exemplary embodiment of the invention may be implemented as part of the training phase of a machine vision system and/or as part of run-time processing, i.e., operation of the machine vision system to inspect sample-objects during the active on-line operating cycle.
REFERENCES:
patent: 4578812 (1986-03-01), Yui
patent: 4700400 (1987-10-01), Ross
patent: 4808987 (1989-02-01), Takeda et al.
patent: 4910694 (1990-03-01), Walther
patent: 4977602 (1990-12-01), Beato
patent: 4980706 (1990-12-01), Someya
patent: 4988984 (1991-01-01), Gonzalez-Lopez
patent: 5008752 (1991-04-01), Van Nostrand
patent: 5125043 (1992-06-01), Karlsson
patent: 5276786 (1994-01-01), Long et al.
patent: 5307167 (1994-04-0
Dutta-Choudhury Paul
Reyzin Igor
Safford Brad
Scola Joe
Cognex Corporation
Hesseltine Ryan J.
Russ Weinzimmer
Wu Jingge
LandOfFree
Method and apparatus for inspection for under-resolved... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and apparatus for inspection for under-resolved..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for inspection for under-resolved... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3194486