Wavelength-Resolution SAR Ground Scene Prediction Based on Image Stack

Abstract

This paper presents five different statistical methods for ground scene prediction (GSP) in wavelength-resolution synthetic aperture radar (SAR) images. The GSP image can be used as a reference image in a change detection algorithm yielding a high probability of detection and low false alarm rate. The predictions are based on image stacks, which are composed of images from the same scene acquired at different instants with the same flight geometry. The considered methods for obtaining the ground scene prediction include (i) autoregressive models; (ii) trimmed mean; (iii) median; (iv) intensity mean; and (v) mean. It is expected that the predicted image presents the true ground scene without change and preserves the ground backscattering pattern. The study indicates that the the median method provided the most accurate representation of the true ground. To show the applicability of the GSP, a change detection algorithm was considered using the median ground scene as a reference image. As a result, the median method displayed the probability of detection of 97 % and a false alarm rate of 0.11 / km 2 , when considering military vehicles concealed in a forest.

CNN-Based Change Detection Algorithm for Wavelength-Resolution SAR Images

“…Still, for the above operating point, Table V compares the P d and FAR of the proposed solution with selected operating points of other CDAs [4], [7]- [9]. The results indicate that the proposed CDA outperforms other state-ofthe-art wavelength-resolution SAR CDAs by a considerable margin, while at the same time requiring the least number of SAR images to make a prediction.…”

“…2 in terms of a receiver operating characteristic (ROC) curve, obtained by varying the segmentation threshold ω 1 . For comparison, the performances of other CDAs tested on the CARABAS-II data set [4], [7]- [9] are also shown. As can be seen, even without the classification step, the proposed method already outperforms the existing ones for all FAR ≥ 0.174/km 2 .…”

“…For instance, two approaches have achieved reasonable detection performances and low false alarm rates (FARs) by using CFAR detection methods considering the use of only two SAR images [4], [6]. On the other hand, we can find interesting change detection approaches that use SAR image stacks [7], [8]. The CDAs presented in [7] and [8] explored the stacks in different ways.…”

“…On the other hand, we can find interesting change detection approaches that use SAR image stacks [7], [8]. The CDAs presented in [7] and [8] explored the stacks in different ways. While the former uses the stacks to obtain a ground scene prediction without targets and uses the result as inputs of a very simple CDA, the latter proposes a CDA based on a noise canceller algorithm, using three SAR images.…”

“…ROC curves of the proposed CDA without the classification step (obtained by varying ω 1 ) and with the classification step (obtained by varying ω 2 for fixed ω 1 = 0.68). Curves or points extracted from[4],[7]-[9] are also shown for comparison.…”