System for detection of non-uniformities in web-based materials

A method for inspecting web material in real-time and computing a severity of a non-uniformity pattern as the web material is manufactured, the method comprising:

• receiving, with an online computerized inspection system, an image captured from a web material being manufactured;
• processing the image with the inspection system to extract a first set of features;
• processing the first set of features with the inspection system with a decision-making model that defines a chain of triggers to dynamically select a second set of features to extract from the image in accordance with the model;
• determining, in accordance with the model, an expected reduction in cost of misclassifying the image with respect to the non-uniformity pattern if the second set of features are extracted from the image;
• when the expected reduction in cost exceeds a threshold, invoking hardware components of a pipelined graphical processing unit (GPU) integrated within the inspection system to process the image with the inspection system to extract the second set of features;
• and
• computing the severity level for the non-uniformity pattern for the web material with the inspection system using the first set of features and the second set of features extracted from the image.

An online computerized inspection system for inspecting web material in real-time and computing a severity of a non-uniformity pattern as the web material is manufactured, the inspection system comprising:

• a memory to store a decision-making model that defines a chain of dynamic triggers for selecting in real-time which features to extract from an image;
• and
• a computer executing software to process a sample image captured from a manufactured web material currently being manufactured to extract a first set of features,
• wherein the software applies the decision-making model to repeated dynamically trigger selection of additional sets of features and to extract from the sample image and invoke hardware components of a pipelined graphical processing unit (GPU) of the inspection system to extract the additional sets of features until the extracted features are sufficient to compute a severity level of a non-uniformity pattern for the web material within a defined level of confidence.

An online computerized inspection system for inspecting web material in real-time comprising:

• a memory to store a model that represents a continuous ranking of the training images as a plurality of points within a multidimensional feature space, wherein each of the points within the multidimensional space corresponds to a feature vector for a different one of the training images;
• and
• a pipelined graphical processing unit (GPU) of the inspection system to process a sample image captured from a manufactured web material to extract a plurality of features in parallel from the sample image using parallel hardware components of the GPU,
• wherein the computerized inspection system computes a severity level of a non-uniform pattern from the extracted features for the sample image of the web material on a continuous scale based on the model of the training images.

A system comprising:

• rating software executing on a processor, wherein the rating software extracts features from each of a plurality of training images by computing a numerical descriptor for each of the training images from pixel values of the respective training image, wherein the rating software performs a first clustering process to process the numerical descriptors of the training images to automatically select a representative subset of the training images and compute a plurality of image clusters for the representative subset of training images;
• and
• a user interface presented by the rating software having input mechanisms to receive input from a user specifying one or more classes of patterns present within the representative training images and a set of rating labels for each of the classes of patterns, wherein the user interface further includes input mechanisms to receive input assigning an individual rating label to each of the image clusters for each of the specified classes of patterns, wherein, for each of the image clusters, the rating software automatically propagates each of the individual rating labels assigned to the classes of patterns for the image cluster to all of the training images within that image cluster;
• training software to compute a classification model based on the rating labels assigned to the training samples;
• and
• an online computerized inspection system comprising a pipelined graphical processing unit (GPU) of the inspection system to process a new image captured from a manufactured web material to extract a plurality of features in parallel from the new image and compute a severity level of a non-uniform pattern for the web material continuous scale based on the classification model.

A method of training a user comprising:

• executing rating software on a computer to automatically assign a discrete rating label for a non-uniform pattern to each of a plurality of training images and compute a classification model based on the rating labels assigned to the training samples by one or more experts;
• processing in real-time a sample image captured from a manufactured web material with a computerized inspection system to extract a first plurality of features from the image;
• computing a severity level for the non-uniformity pattern for the sample image of the web material with the computerized inspection system from the first plurality of features in accordance with the classification model;
• and
• displaying both the sample image and the severity level computed from the classification model to a user as training information for visualizing the non-uniformity pattern in the same manner as the experts.