Cross-resolution difference learning for change detection between multitemporal images

3 years ago 327
Researchers suggest  cross-resolution quality  learning for unsupervised alteration  detection Flowchart of the projected CD method. Credit: XIOPM

Recently, a squad led by Prof. Lu Xiaoqiang from the Xi'an Institute of Optics and Precision Mechanics (XIOPM) of the Chinese Academy of Sciences projected cross-resolution quality learning for unsupervised alteration detection. Their up-to-date effect was published successful IEEE Transactions connected Geoscience and Remote Sensing.

Change detection (CD) chiefly aims astatine recognizing the differences betwixt multitemporal images captured implicit the aforesaid geographical country astatine antithetic times. Compared with methods based connected cumbersome labeled alteration , unsupervised CD methods tin make a alteration representation without astir the alteration information, which has attracted wide attention.

Moreover, it is hard to straight observe changes successful the applicable application, due to the fact that galore multitemporal images captured astatine antithetic times person antithetic resolutions with antithetic sensor properties. For astir existing methods, they usually resized multitemporal images to a unified size which has a connected the last CD show due to the fact that of changing the archetypal accusation of pixels.

To code the supra problems, Lu and his projected a cross-resolution quality learning method without resizing operations and cumbersome labels. The full model was disassembled into 3 modules, representation segmentation, quality learning, and cross-resolution fusion.

According to the experiments results, the effectiveness of the projected method are demonstrated nether antithetic valuation metrics. In the future, the projected CD method volition supply a usher for designing caller model of cross-resolution unsupervised alteration detection.



More information: Xiangtao Zheng et al, Unsupervised Change Detection by Cross-Resolution Difference Learning, IEEE Transactions connected Geoscience and Remote Sensing (2021). DOI: 10.1109/TGRS.2021.3079907

Citation: Cross-resolution quality learning for alteration detection betwixt multitemporal images (2021, August 30) retrieved 30 August 2021 from https://techxplore.com/news/2021-08-cross-resolution-difference-multitemporal-images.html

This papers is taxable to copyright. Apart from immoderate just dealing for the intent of backstage survey oregon research, no portion whitethorn beryllium reproduced without the written permission. The contented is provided for accusation purposes only.

Read Entire Article