Image correlation for end-to-end displacement and strain measurement

Patent metadata

Field	Value
Publication number	US-12540815-B2
Application number	US-202217953033-A
Country	US
Kind code	B2
Filing date	Sep 26, 2022
Priority date	Sep 27, 2021
Publication date	Feb 3, 2026
Grant date	Feb 3, 2026

Abstract

A system for correlating image data includes a memory configured to store a sequence of images of a sample. The system also includes a processor operatively coupled to the memory and configured to crop a first pair of images to specify a region of interest in the first pair of images, where at least one image in the pair of images is from the sequence of images. The processor is also configured to calculate, using a first convolutional neural network, a displacement field for the first pair of images. The processor is also configured to calculate, using a second convolutional neural network, a strain field for the first pair of images. The processor is further configured to determine an amount of displacement or deformation of the sample based at least in part on the displacement field and the strain field.

First claim

What is claimed is:

1 . A system for correlating image data, the system comprising:
a memory configured to store a sequence of images of a sample; a processor operatively coupled to the memory and configured to: crop a first pair of images to specify a region of interest in the first pair of images, wherein at least one image in the pair of images is from the sequence of images; calculate, using a first convolutional neural network, a displacement field for the first pair of images; calculate, using a second convolutional neural network, a strain field for the first pair of images; and determine an amount of displacement or deformation of the sample based at least in part on the displacement field and the strain field, wherein the processor is configured to use the displacement field to generate two image outputs, and wherein each of the two image outputs has a size of h×w.

2 . The system of claim 1 , wherein the strain field is calculated independent of the displacement field.

3 . The system of claim 1 , wherein the first pair of images includes a reference image and a deformed image, wherein the deformed image is a deformed version of the reference image.

4 . The system of claim 3 , wherein the processor generates the deformed image by warping the reference image.

5 . The system of claim 1 , wherein the processor is configured to determine an updated region of interest based at least in part on the calculated displacement field.

6 . The system of claim 5 , wherein the processor is configured to determine the updated region of interest based on updated coordinates of four corner points in the displacement field such that the updated region of interest tracks a deformation of the sample.

7 . The system of claim 5 , wherein the processor is configured to crop a second pair of images using the updated region of interest.

8 . The system of claim 7 , wherein the processor is further configured to:
calculate, using the first convolutional neural network, an updated displacement field for the second pair of images; determine a subsequent updated region of interest based at least in part on the updated displacement field for the second pair of images; and crop a third pair of images using the subsequent updated region of interest.

9 . The system of claim 1 , wherein the system is trained with one or more synthetic datasets.

10 . The system of claim 1 , wherein the processor is configured to use the strain field to generate three image outputs, wherein each of the three image outputs has a size of h×w, and wherein each of the three image outputs includes a plane strain component.

11 . A method for correlating image data, the method comprising:
storing, in a memory of a computing system, a sequence of images of a sample; cropping, by a processor operatively coupled to the memory, a first pair of images to specify a region of interest in the first pair of images, wherein at least one image in the pair of images is from the sequence of images; calculating, by the processor and using a first convolutional neural network, a displacement field for the first pair of images; calculating, by the processor and using a second convolutional neural network, a strain field for the first pair of images; determining, by the processor, an amount of displacement or deformation of the sample based at least in part on the displacement field and the strain field; and generating, by the processor using the displacement field, two image outputs, wherein each of the two image outputs has a size of h×w.

12 . The method of claim 11 , wherein calculating the strain field comprises calculating the strain field independent of the displacement field.

13 . The method of claim 11 , wherein the first pair of images includes a reference image and a deformed image, and further comprising forming the deformed image by warping the reference image.

14 . The method of claim 11 , further comprising determining, by the processor, an updated region of interest based at least in part on the calculated displacement field.

15 . The method of claim 14 , wherein the processor is configured to determine the updated region of interest based on updated coordinates of four corner points in the displacement field such that the updated region of interest tracks a deformation of the sample.

16 . The method of claim 14 , further comprising cropping, by the processor, a second pair of images using the updated region of interest.

17 . The method of claim 16 , further comprising:
calculating, by the processor and using the first convolutional neural network, an updated displacement field for the second pair of images; determining, by the processor, a subsequent updated region of interest based at least in part on the updated displacement field for the second pair of images; and cropping, by the processor, a third pair of images using the subsequent updated region of interest.

18 . The method of claim 11 , further comprising training the system with one or more synthetic datasets.

19 . The method of claim 11 , further comprising:
generating, by the processor using the strain field, three image outputs, wherein each of the three image outputs has a size of h×w, and wherein each of the three image outputs includes a plane strain component.

Assignees

• Univ Northwestern

Inventors

• Guo Ping
• Yang Ru

Classifications

• G06T2207/20164

  Salient point detection; Corner detection · CPC title

• G06T2207/20132

  Image cropping · CPC title

• G06T2207/20081

  Training; Learning · CPC title

• G06T2207/10016

  Video; Image sequence · CPC title

• G06V10/82

  using neural networks · CPC title