Estás aquí: start » hiperespectral » cva

GPU Framework for Change Detection of Multitemporal Hyperspectral Images

Esta é unha revisión antiga do documento!


Experimental results related to the paper GPU-based Change Detection Applied to Multitemporal Agricultural Hyperspectral Images.

Abstract

Nowadays, it is increasingly common to have access to several multidimensional remote sensing images captured in the same region at different timeframes which allows the study of the changes over these areas. The biggest part of the available change detection techniques are focused on pixel-based operations. In this work, a binary spectral-spatial change detection technique to find object-based differences in multitemporal remote sensing hyperspectral data is presented. The technique is based on Change Vector Analysis (CVA) improved with Spectral Angle Mapper (SAM) distance and Otsu's thresholding. It also involves a initial watershed segmentation-based technique to take into account the spatial information of the images. The technique can be efficiently executed in real-time applications thanks to its projection to GPU devices.

Downloads

Input datasets

For information see the readme files in the archives.

* Barbara

* Bay Area

Results

For information see the readme file in the archives.

*outputscva.zip

License