Methods and systems for processing images for inspection of an object
AssigneeGeneral Electric Company
InventorsGlen William Brooksby, Paulo Ricardo Dos Santos Mendonca, Shubao Liu, Qi Song
Life of this patent
Estimating the term…
Abstract
Methods and systems for processing images are provided. One method includes obtaining a plurality of images corresponding to at least one area of an object and performing a rectification of at least some of the plurality of images using a reference structure. The method also includes performing a gradient vector field analysis on the rectified plurality of images of the object to identify anomaly regions within the object.
Claims19
A non-transitory computer readable storage medium for Identifying anomalies in an object that is imaged using at least one processor, the non-transitory Computer readable storage medium including instructions to command the processor to:
- obtain a plurality of images corresponding to at least one area of an object;
- perform a rectification of at least some of the plurality of images using a reference structure;
- and
- perform a gradient vector field analysis on the rectified plurality of images of the object to identify anomaly regions within the object, wherein the object includes a membrane and the instructions command the processor to identify the membrane using a model of the membrane, the membrane is identified in the plurality of images by optimizing a cost function that assesses fitness to at least one of the model or agreement with image data, and the cost function is defined as:
- C
-
- (
- S
- )
- =
- C
- intensity
-
- (
- S
- )
- +
- ∑
- S
- adj
-
- (
- S
- -
- S
- adj
- )
- 2
- +
- (
- [
- S
- -
- S
- model
- )
- ]
- 2
- wherein CIntensity(S) is a measure of the fitness of a first or a current membrane detection S with respect to a defined intensity model of the membrane, Sadj is a current membrane detection in adjacent images and Smodel is a prior shape model.
An inspection system comprising:
- a processor configured to perform a rectification of at least some of a plurality of images of an object using a reference structure and to perform a gradient vector field analysis on the rectified images of the object to identify anomaly regions within the object;
- and
- a display configured to display at least one of the images with any identified anomaly regions on the displayed image, wherein the object is a propeller blade that includes an internal membrane and the processor is further configured to identify the membrane using a global cost function summed across the plurality of images, wherein the global cost function is defined as:
- C
-
- (
- S
- )
- =
- C
- intensity
-
- (
- S
- )
- +
- ∑
- S
- adj
-
- (
- S
- -
- S
- adj
- )
- 2
- +
- (
- [
- S
- -
- S
- model
- )
- ]
- 2
- wherein CIntensity(S) is a measure of a fitness of a first or a current membrane detection S with respect to a defined intensity model of the membrane, Sadj is a current membrane detection in adjacent images and Smodel is a prior shape model.
A method for automatically identifying anomalies in images of an object having an internal membrane, the method comprising:
- identifying the internal membrane in a plurality of images of the object;
- determining an overall shape of the object;
- warping the overall shape in the images to straighten the object;
- determining local standard deviations of image intensities to identify material types within the object;
- computing and regularizing intensity gradients across a region of interest within the plurality of images to determine a gradient vector field;
- applying a curl operator to the gradient vector field;
- thresholding a result of the curl operator to determine any anomaly region;
- applying a region growing process to any anomaly region;
- and
- generating and displaying an overlay on at least one of the plurality of images identifying any anomaly region.
Dependencies are inferred from the claim text.