Change Detection

Change Detection Using SAR Images

Under the support of ONR, we recently implemented a change detection algorithm using SAR images. Our algorithm compares two images (one reference and one test), which are filtered first. After that, a residual is generated using our sparsity based change detection algorithm based on partial frames. The residual is then thresholded to generate the final detection results. The information flow is shown in the following figure.

Change Detection Using Hyperspectral Images

This work was supported by AFOSR. Hyperspectral images were supplied from AF Wright Patterson Lab. Here the reference image was collected in August and the test image was collected in October. We developed a sparsity based change detection algorithm based on partial frames and the detection results are shown in (c). The ROC curves of two other methods are also shown in (d).


MRCDTM is an advanced change detection framework which can succesfully handle illumination change, seasonal change, parallax and occlusion. The idea is to incorporate multiple reference images for change detection. Here we include one example. It can be seen that our MRCD has much fewer false alarms than a conventional method.

Original Images. Left: One of the reference images; right: Test image.

Detection Results. Left: A conventional method: CC-CKX; right: Our method: MRCD.

Contact: or 240-505-2641

Copyright © 2015 Signal Processing, Inc. All rights reserved.