Method and Product for Detecting Abnormalities
AssigneeABB AB
InventorTommi Huotilainen
Life of this patent
Estimating the term…
Abstract
A new method for processing image data in order to detect abnormalities in a web is provided. The web is monitored by at least one camera, whereby an image comprising of plurality of pixels is generated. The data of the image is stored in a memory. Image data is filtered by a processor by creating a filtered image data by weighting the image data and at least one of earlier image data and earlier filtered image data; and combining the weighted image data and at least one of the weighted earlier image data and the weighted earlier filtered image data; and controlling filtering by at least one nonlinear algorithm; and thresholding the created filtered image data.
Claims29
A method for processing image data in order to detect abnormalities in a web, wherein the web is monitored by at least one camera, whereby at least one image comprising of plurality of pixels is generated for creating the image data, and the image data is stored in a memory, wherein the image data (x(n)) is filtered by a processor for creating a filtered image data (y(n)), wherein the current filtered image data (y(n)) is obtained by weighting the current image data (x(n), a0, b0) and at least one of image data (x(n)) from emphasis angles and filtered image data (y(n)) from emphasis angles;
- and by combining the obtained weighted current image data and the obtained at least one of weighted image data from emphasis angles and weighted filtered image data from emphasis angles, wherein the emphasis angles are dependent on the result of previous filtering;
- and wherein the filtering of the current image data (x(n)) is controlled by at least one nonlinear algorithm;
- and followed by thresholding the created filtered image data (y(n)).
A computer program product stored on computer operable media for processing image data in order to detect abnormalities in a web, wherein the web is monitored by at least one camera, whereby at least one image comprising of plurality of pixels is generated for creating the image data (x(n)), and the image data (x(n)) is stored in a memory, wherein the image data (x(n)) is filtered by a processor for creating a filtered image data (y(n)), wherein the current filtered image data (y(n)) is obtained by weighting the current image data (x(n), a0, b0) and at least one of image data (x(n)) from emphasis angles and filtered image data (y(n)) from emphasis angles;
- and by combining the obtained weighted current image data and the obtained at least one of weighted image data from emphasis angles and weighted filtered image data from emphasis angles, wherein the emphasis angles are dependent on the result of previous filtering;
- and wherein the filtering of the current image data (x(n)) is controlled by at least one nonlinear algorithm;
- and followed by thresholding the created filtered image data (y(n)).
Dependencies are inferred from the claim text.