Change detection is a key methodology in remote sensing. Despite numerous important papers dedicated to this topic no comprehensive review of the subject currently exists. Most existing reviews are limited to certain fields of application or are simply no longer current. Here, we provide an overview of the most important algorithms used for change detection and their development and refinement through time. Change detection cannot be considered as a mere algorithm. We thus describe all of the steps necessary to perform change detection as part of a process chain. We also show how such a change detection process chain may be adapted to specific individual requirements. The labeling of changes is recognized as a basic part of this process. Three change labeling categories are introduced: (1) pre-change extraction labeling, (2) concurrent labeling, and (3) post-change extraction labeling. Methods developed specifically for use with synthetic aperture radar (SAR) data are a focus of this review. SAR data provide information compatible with data produced by optical sensors, but also often require specialized processing techniques and are useful within a unique field of application. An examination of time series analysis methods is also included in this review. To date, these techniques have not been considered in reviews, but the increasing availability of remote sensing data as well as recent advances in remote sensing change detection make it essential that they are included here. Although not exhaustive, this review is intended to provide a comprehensive overview of well established change detection methods as well as recent advances in this field.

Recent Advances in Remote Sensing Change Detection – A Review

In this paper we present an alternative method for SAR image denoising, structure enhancement, and change detection based on the curvelet transform. Curvelets can be denoted as a two dimensional further development of the well-known wavelets. The original image is decomposed into linear ridge-like structures, that appear in different scales (longer or shorter structures), directions (orientation of the structure) and locations. The influence of these single components on the original image is weighted by the corresponding coefficients. By means of these coefficients one has direct access to the linear structures present in the image. To suppress noise in a given SAR image weak structures indicated by low coefficients can be suppressed by setting the corresponding coefficients to zero. To enhance structures only coefficients in the scale of interest are preserved and all others are set to zero. Two same-sized images assumed even a change detection can be done in the curvelet coefficient domain. The curvelet coefficients of both images are differentiated and manipulated in order to enhance strong and to suppress small scale (pixel-wise) changes. After the inverse curvelet transform the resulting image contains only those structures, that have been chosen via the coefficient manipulation. Our approach is applied to TerraSAR-X High Resolution Spotlight images of the city of Munich. The curvelet transform turns out to be a powerful tool for image enhancement in fine-structured areas, whereas it fails in originally homogeneous areas like grassland. In the change detection context this method is very sensitive towards changes in structures instead of single pixel or large area changes. Therefore, for purely urban structures or construction sites this method provides excellent and robust results. While this approach runs without any interaction of an operator, the interpretation of the detected changes requires still much knowledge about the underlying objects.

/pdf/curvelet-approach-for-sar-image-denoising-structure-3o35efabn7.pdf

Curvelet approach for sar image denoising, structure enhancement, and change detection

This article compares four different alternative image representations in the context of a structure-based change detection The framework is taken from the already published Curvelet-based change detection approach Only the transform step is modified by inserting three additional transforms: the Laplacian pyramid, the Wavelet and the Surfacelet transform The results of the change detection are compared to the single pixel difference image in order to find the representation that best illustrates the underlying structures The Curvelet transform again turns out to be very powerful in describing man-made objects and landscapes

/pdf/comparison-of-alternative-image-representations-in-the-yr3awe2c2x.pdf

Comparison of alternative image representations in the context of SAR change detection

Autoencoders are a type of neural network that reproduce in the ideal case the input signal at the output. In an intermediate step of these networks, a code, which should contain all relevant information of the input signal, is created. In this work, an average code for the background of the observed area is created and deviations from this background code are used to detect targets in the observed region. The Autoencoder is trained with available background data and additionally with profiles created by noise in the dynamic range of the radar. Results are shown for range profiles measured with a 94GHz radar.

Target Detection using Autoencoders in a Radar Surveillance System

In this paper the methods and operational constraints of change detection from synthetic aperture radar images are reviewed. A simulation-based environment dedicated to the evaluation of a change detection algorithm is described. Some results are presented for a linear feature detection algorithm.

L. Polidori

Papers

Change detection in radar images: methods and operational constraints