Medar, Inc.

Integral Vision, Inc.

David A Robinson

Anthony R Bland

A method and system are described for use at a machine vision station to correctly identify flaws within a sample image by means of flexible template image comparison where the template image and a sample image suffer from a relative geometrical distortion. The method and system use the following iterative approach: 1) generate the template image; 2) register the sample image to the template image; 3) undertake an absolute image subtraction process; 4) identify all regions of image difference; 5) create sub-images of both the template and the sample images in the region of detected error; 6) repeat the registration step 2) using the sub-images; 7) undertake the absolute subtraction process on each of the sub-images; and 8) either identify the resulting differences as flaws or go to step 5) using even smaller sub-images. The invention has particular utility in the screen printing industry as well as other industries.

Method and system for detecting a flaw in a sample image

A method and system are provided for identifying an image feature of an object. The system preferably includes a machine vision system for capturing an original color image of the object in a first color space defined by a first set of color axes to obtain a first set of image data. The first set of image data is transformed within the machine vision system to a second set of image data in an optimal color space having an optimal set of color axes which define an optimal direction wherein the amount of inter axis correlation of the second set of image data is minimized. The machine vision system then manipulates the second set of image data to identify the image feature. Color differences existing in the original color image are amplified and separation between chromatically different regions of the original color image are maximized in the optimal color space. Preferably, the optimal set of color axes are three orthogonal axes. The first set of image data is rotated to match the optimal direction. In this way, the first set of image data is reorganized along the optimal set of color axes. The machine vision system can also be utilized to capture a color image of an ideal object in the first color space to obtain an ideal set of image data and to compute the optimal set of color axes from the ideal set of image data wherein the optimal set of color axes is object specific.

Method and system for identifying an image feature and method and system for determining an optimal color space for use therein

A calibration algorithm is provided for calibrating a machine vision system including a color camera, a framegrabber and an illumination source wherein subsequently acquired images are processed on a pixel-by-pixel basis. After framegrabber offsets are measured to obtain initial or offset calibration data, the camera is pointed at a target monotone patch of known reflectance such as pure white and the camera outputs for red, green and blue are determined while observing the monotone patch to obtain spatial calibration data. The gain for each of the output channels of the color camera is automatically determined from the initial calibration data. Applying the three gain factors to a subsequent image on the pixel-by-pixel basis allows one to end up with a color-corrected image without incurring any time performance penalty. At any time during an inspection process, a small area of white reflectance can be observed. The resulting dynamic white calibration data is used to correct for any medium to long term temporal variation in imaging conditions.

Method and system for automatically calibrating a color-based machine

A method and system are described for use at a machine vision station to correctly identify errors within a sample image of an object by means of a probability image where a value of each pixel of the probability image is a probability. The probability image is created from a mean template image and a standard deviation image, both of which are created from a plurality of reference images of the object. The probability image is processed to detect the errors in the sample image independent of the values of the pixels which make up the sample image. Two different techniques are described for creating the probability image utilizing mean template, sample and standard deviation pixel values.

Method and system for detecting errors in a sample image

A pair of calibration algorithms are provided for calibrating a color camera in a machine vision system. The first and most general form of the calibration algorithm requires that a calibration target be shown to a well adjusted machine vision system. The calibration target contains samples of red, green and blue. A gray level histogram is taken of each of the resulting red, green and blue images and is used to generate corresponding standard cumulative probability functions. The standard cumulative probability functions, together with a copy of the calibration target, is provided to a machine vision system having its own color camera which requires calibration. The same above-noted steps are repeated for this system and the resulting cumulative probability functions are compared to the corresponding standard cumulative probability functions to obtain gains for the red, green and blue channels of the camera to be calibrated . In the second simplified version of the calibration algorithm, a target patch of pure white is provided and camera outputs for red, green and blue are determined while observing the patch. The average gain for each of the output channels of the color camera is determined. This simplifies to taking the average white level measured for each of the channels and dividing the average white value into the full scale value of the machine vision system to be calibrated. Applying the three gain factors determined under either version of the algorithm to a subsequent image allows one to end up with a color-corrected image.

Method and system for automatically calibrating a color camera in a

Color balance of a sample image is compared with a previously learned template image utilizing an appropriately programmed machine vision system including a color camera at a vision station. The sample image, preferably captured from a web scanner attached directly to an offset printing machine, is compared with a previously generated 'known good' template image. This comparison process produces four correction factors which are the ratios between the amount of each of four colored inks in the template image to that in the sample image. The four colored inks may be cyan, magenta, yellow, and black inks but inks of other colors could also be used. These correction factors can either be presented to an operator to manually adjust the machine or else used in an automatic feedback loop of a controller of the machine, thereby maintaining consistent image quality.

Method and system for automatically monitoring the colors of an object

A method and system are provided for automatically and quantitatively measuring color difference between a color distribution of an object and a reference color image using 'color distance' in a color system. A template representing the reference color image is stored in a memory of a machine vision system. The machine vision system generates a sample color image of the object and processes the template together with the sample color image to obtain a total color distance. Color distance may be measured either in a Euclidean or 'City Block' fashion. Preferably, a look-up table stores the template. The color system may be either intensity dependent or independent. The total color distance can either be presented or displayed to an operator of the machine vision system or else can be used in an automatic feedback loop of a controller of a machine or process.

Growing community of inventors

Northants, United Kingdom

David A Robinson