Rotation invariant change-detection
PC: https://www.azavea.com/blog/2022/04/18/change-detection-with-raster-vision/

Rotation invariant change-detection

Problem: Change detection is a much researched area. Most algorithms run on temporal satellite images. Case in point is when the source of images are different (say for instance trying to compare batches of UAV fed images to satellite images of a particular area) - where the orientation or resolution or both cannot be expected to be the same. A typical military use case is to detect change in deployment along a large area.

Solution: SURF/SIFT algorithms detect 'features' and are rotation invariant. Typical use cases - similarities between two images. The same features can be used to detect changes too.

How? Apply the algorithm (SIFT/SURF) to pick features in both the images. Compare features to select ones that are unique in each image.

No alt text provided for this image
Same orientation and resolution : Unmatched Features detected on Galwan Valley satellite images of 22 Jun 2020(old image) and 25 Jun 2020 (new image)
No alt text provided for this image
Images from different heights. Left image - a close up image of the area. Right image covering a larger area. There is a common region in both the images.
No alt text provided for this image
Matched region highlighted
No alt text provided for this image
Deployment in the left image are marked as unmatched features (which are absent on the right image).
G Singh

Geospatial Technologies & Remote Sensing | Innovator in Generative AI Applications | Holder of Multiple Patents & Renowned International Researcher | Advanced Geoinformatics & Technical Expert

2 年

Call me . We can discuss There are certain other robust ways as well to solve this particular military problem

Ramneek Kapoor

Security Management || Operations Management || People And Processes || Indian Army Veteran

2 年

Complex Problem.... Simple Solution... Great Job ??

回复
Tresa Mary Joseph

JP Morgan Chase & Co. | Military Veteran

2 年

That's a lot to do with How? & Why of many image ?? interpretations and conclusions. Best wishes on your studies ?. Great going ??

回复

要查看或添加评论,请登录

Vinay Chandran的更多文章

  • One shot detector or two shot detector?

    One shot detector or two shot detector?

    While many comparative studies on two shot object detectors and one shot object detectors wherein most of them claiming…

社区洞察

其他会员也浏览了