Practical approach for synthetic aperture radar change analysis in urban environments

Abstract

Change detection using remote sensing imagery is a broad and highly active field of research that has produced many different technical approaches for multiple applications. The majority of these approaches have in common that they do not deliver any detailed information concerning the type, category, or class of the detected changes. With respect to the extraction of such information, recent research often suggests that a land use classification is required. This classification can be accomplished in an unsupervised or supervised way, whereas the practicability of both strategies is more or less limited by the usage of reference or training data. Moreover, expert knowledge is needed to arrive at meaningful land use classes. An approach is presented that overcomes these drawbacks. A time series of synthetic aperture radar amplitude images is considered, enabling the detection of so-called high activity objects in urban environments. Such objects represent the basis of the investigations and denote the input for unsupervised categorization and classification procedures. The method supports even the unexperienced user in learning the actual information content leading to the capability to define a suitable scheme for change classification. Tests carried out on two different datasets suggest that the method is both practical and robust. © The Authors. Published by SPIE under a Creative Commons Attribution 4.0 Unported License. Distribution or reproduction of this work in whole or in part requires full attribution of the original publication, including its DOI.

Training data generation for machine learning using GPR images

“…This background image is further be used as reference for the change detection step, in which the other images of the sequence are analyzed. The change detection method itself is applied as described in [9], where it was successfully used to find changes in SAR amplitude images. As result from the change detection step, a binary image exists, in which the changes, or objects of interest (OOI), between the mean background image and the specific image of the sequence are placed in the foreground.…”

“…To produce result images, which are visually enhanced and free of noise that possible makes it hard for the operator to interpret the image contents manually, a morphological filtering of the mean background image is performed. For this step, an Alternating Sequential Filter (ASF) by considering the area attribute is applied [9]. Finally, the original signature of the OOI is imported to the ASF filtered background image.…”

Classification of Remote Sensing Data With Morphological Attribute Profiles: A decade of advances

“…While the application of APs to optical remote sensing data has been strongly focused on, alternative remote sensing image types have received far less attention. One may witness some tentative works on SAR (Synthetic Aperture Radar) and polarimetric SAR images for segmentation [33], building detection [34], crop field and land-cover classification [35], [36] and change detection [37]- [39] using the original APs and the Differential Attribute Profiles; on passive microwave remote sensing image analysis [40]; on LiDAR data for building detection [41] and land cover classification [16], [28], [42]- [44]; on satellite image time-series classification using Sentinel-2 data [45], [46]; on the fusion of APs and Extinction Profiles (a variant of AP that will be discussed in Sec. III-F) of hyperspectral and LiDAR data using composite kernel SVM [47], [48] and deep learning approaches [49], [50] for land cover classification.…”

“…Although the aforementioned four attributes are by far the most widely encountered in the state of the art, APs can accommodate (from a theoretical point of view) a vast pool of attributes. Examples include entropy, homogeneity [7], as well as the diameter of equivalent circle and area of convex hull for automatic threshold selection [64]; complexity (perimeter over area) [65]; perimeter and area of bounding box used to evaluate threshold-free APs [66]; solidity (area over area of convex hull) and orientation (between the major axis of the convex hull and the x-axis) [67]; Cov (Coefficient of variation) and NRCS (Normalized Radar Cross Section) tailored for SAR images [33], [39], where Cov is the ratio of the standard deviation divided by the mean value of pixel intensities, and NRCS, expressed in decibels, is the radar cross section per unit area of surface. Furthermore, in [68], it has been observed that when dealing with multiband input, one can extend the pool of attribute measures to include multi-dimensional functions exploiting all available bands simultaneously and two new attributes have been proposed: higher-dimensional spread and dispersion.…”