Satellite imagery

Patent metadata

Field	Value
Publication number	US-12322121-B2
Application number	US-202117227413-A
Country	US
Kind code	B2
Filing date	Apr 12, 2021
Priority date	Apr 28, 2020
Publication date	Jun 3, 2025
Grant date	Jun 3, 2025

Abstract

A system for detecting changes between two temporally different images includes an image divider, a Convolutional Neural Network (CNN) feature encoder, an image alignment system, a feature comparator, a CNN feature decoder and segmenter, and a block combiner. The image divider divides a first and second image into a plurality of image blocks. CNN feature encoder encodes the image blocks from the first and second image into first and second feature sets respectively. The image alignment system aligns the first and second image by searching for matching anchor vectors in the first and second feature sets using a similarity search. The feature comparator produces change feature sets from the first and second feature sets of the aligned image blocks, and the CNN feature decoder and segmenter creates segmented change image blocks from the change feature sets. The block combiner combines segmented change image blocks into a segmented change image.

First claim

What is claimed is:

1. A system for detecting changes between two temporally different images, the system comprising:
an image divider to use a grid to divide a first satellite image into a plurality of first satellite image blocks, and to divide a second satellite image, which is spatially similar but temporally different from said first satellite image, into a plurality of second satellite image blocks;
a Convolutional Neural Network (CNN) feature encoder to encode said first satellite image blocks into first block feature sets, and to encode said second satellite image blocks into second block feature sets;
an image alignment system to identify as pairs of anchor blocks those first block feature sets and second block feature sets which match within a predefined threshold according to a similarity search and to use a plurality of said pairs of anchor blocks and their locations within said first satellite image and said second satellite image to align said first satellite image and said second satellite image;
a feature comparator to produce change feature sets from said first and second block feature sets of said aligned image blocks;
a CNN feature decoder and segmenter to create segmented change image blocks from said change feature sets; and
a block combiner to combine a plurality of said segmented change image blocks into a segmented change image.

2. The system of claim 1 , wherein said similarity search is a K nearest neighbor search.

3. The system of claim 2 wherein said similarity search to use one of: Euclidian, cosine, Hamming, and L1 distance metrics.

4. The system of claim 1 wherein said change feature sets comprise those of said second block feature sets where changes between said first satellite image and said second satellite image exist.

5. The system of claim 1 , said feature comparator to operate on block feature sets of non-anchor image blocks.

6. A method for detecting changes between two temporally different images, the method comprising:
using a grid to divide a first satellite image into a plurality of first satellite image blocks, and dividing a second satellite image, which is spatially similar but temporally different from said first satellite image, into a plurality of second satellite image blocks;
encoding said first satellite image blocks into first block feature sets, and encoding said second satellite image blocks into second block feature sets;
identifying as pairs of anchor blocks those first block feature sets and second block feature sets which match within a predefined threshold according to a similarity search;
aligning said first satellite image and said second satellite image using said pairs of anchor blocks and their locations within said first satellite image and said second satellite image;
encoding a first correlated image block into a first block feature set, and encoding a second correlated image block into a second block feature set;
producing a change feature set from said first and second block feature sets of said aligned image blocks;
decoding and segmenting said aligned image blocks to create a segmented change image block from said change feature set; and
combining a plurality of said segmented change image blocks into a segmented change image.

7. The method of claim 6 , wherein said similarity search is a K nearest neighbor search.

8. The method of claim 7 , wherein said similarity search to use one of: Euclidian, cosine, Hamming, and L1 distance metrics.

9. The method of claim 6 wherein said change feature set comprise those of said second block feature sets where changes between said first satellite image and said second satellite image exist.

10. The method of claim 6 wherein said producing is performed on block feature sets of non-anchor image blocks.

Assignees

• Gsi Technology Inc

Inventors

• Erez Elona
• Akerib Avidan

Classifications

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/0455

  Auto-encoder networks; Encoder-decoder networks · CPC title

• G06V20/13

  Satellite images · CPC title

• G06V10/50

  by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis · CPC title

• G06V10/82

  using neural networks · CPC title