Data processing: measuring – calibrating – or testing – Measurement system – Statistical measurement
Reexamination Certificate
1999-03-05
2002-01-01
Wachsman, Hal (Department: 2862)
Data processing: measuring, calibrating, or testing
Measurement system
Statistical measurement
C345S419000, C340S658000, C382S110000
Reexamination Certificate
active
06336082
ABSTRACT:
BACKGROUND OF THE INVENTION
The invention relates to a method and apparatus for screening n-dimensional data for abnormalities. One application of the invention is inspecting x-ray images of objects to detect abnormalities. Currently, x-ray inspection is performed on manufactured parts (e.g. turbine blades) to detect defects in the item. Conventional flaw detection algorithms use image processing techniques such as local fitting, convolution, or morphological operators to enhance flaws by comparing a center pixel against its neighbors. These local operators not only enhance flaws but also enhance sharp profiles such as edges of the part that are not flaws. Accordingly, the enhanced images need to be inspected visually by human operators to differentiate between a flaw occurrence and the correct part geometry. Human operators are prone to eye fatigue, and thus reduced accuracy, and fast turn over rate.
Automated visual inspection of blades in aircraft and power generation turbines must be exhaustive, fast, and flexible. Currently, this inspection task is done by human inspectors looking at 16-bit x-ray images after the images have been enhanced with a high-pass filter. Previous approaches to automatic flaw detection generally use local operators tuned to certain shapes and sizes of the flaws. The operators could be linear convolution, combination of Zernike polynomials, morphological operator, or neural network. All of these approaches detect flaws by comparing the center pixel against its neighboring pixels. This results in enhanced images that have to be visually inspected by humans because the enhancement highlights not only flaws but also other local geometries of the inspected part that have similar spatial characteristics.
Subtracting the image under test from the reference image has been used to detect flaws in printed circuit boards, integrated circuits, and x-ray images of cast parts. A bandpass filter is applied to the images and a flexible matching technique is used to warp the test image before its subtraction from the reference image. Such image subtraction requires extremely accurate registration of the two images before subtraction, and still leaves strong differences at edges and places of large tolerances.
BRIEF DESCRIPTION OF THE INVENTION
An exemplary embodiment of the invention is directed to a method of detecting abnormalities including a learning phase and an inspection phase. In the learning phase, a first statistical quantity and a second statistical quantity are determined based on a first set of n-dimensional data. A second set of n-dimensional data is obtained and a bimodal distribution for a plurality of n-dimensional data points in each of the second set of n-dimensional data is determined. A reference template is generated in response to the first statistical quantity, second statistical quantity and bimodal distribution. In the inspection phase, a test set of n-dimensional data to be inspected for abnormalities is obtained and compared to the reference template to detect abnormalities.
REFERENCES:
patent: 5060290 (1991-10-01), Kelly et al.
patent: 5150111 (1992-09-01), Fujieda et al.
patent: 5546475 (1996-08-01), Bolle et al.
patent: 5668631 (1997-09-01), Norita et al.
patent: 5671344 (1997-09-01), Stark
Mundy Joseph Leagrand
Nguyen Van-Duc
Duong Khoi
General Electric Company
Ingraham Donald S.
Testa Jean K.
Wachsman Hal
LandOfFree
Method for automatic screening of abnormalities 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 for automatic screening of abnormalities, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method for automatic screening of abnormalities will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2816425