SAR Image Change Detection Based on Nonlocal Low-Rank Model and Two-Level Clustering

Abstract

Change detection (CD) has found a wide range of applications in many fields. In this article, we propose a novel nonlocal low-rank (NLR) based method for multitemporal synthetic aperture radar image CD. This method jointly exploits the powerful NLR-based despeckling and the effective cascade clustering. First, the NLR model is used to generate the difference image (DI), which consists of a patch grouping process and a low-rank minimizing process. Especially, the NLR minimization model contains a data fidelity term, which is based on the statistical distribution of speckle noise, and a regularization term, which uses the weighted nuclear norm. Then, the alternating direction methods of multipliers is introduced to solve this minimization problem. Second, after DI is generated, the principal component analysis is employed to extract the feature and a two-level clustering method is used to generate the final change map, which separates the intermediate class by using the neighbor information with Gaussian weighted distance. Experiment results demonstrate the effectiveness of the proposed method by comparing with some state-of-the-art methods.

Change Detection Using Time and Look Angle Series of Geocoded Tandem-X Staring Spotlight Images

“…The UTM images have a fixed pixel spacing in Northing and Easting coordinates with a common south west corner, hence, the RGB (Red, Green, Blue) images can simply be generated by finding the same easting and northing coordinate in the R, G and B image. The simplest way and the most used method to co-register slant range SAR images with almost equal incidence angles is to cross correlate the images [12]. This method doesn't need any accurate orbit data, however, images with large regions with low contrast may be a problem.…”

Hybrid Variability Aware Network (HVANet): A Self-Supervised Deep Framework for Label-Free SAR Image Change Detection

“…Benefiting from the capability of all-weather and all-time Earth observation, the synthetic aperture radar (SAR) sensor has been used in numerous applications increasingly, including but not limited to urban planning, disaster monitoring, and land-cover/landuse (LCLU) analysis [1][2][3][4][5][6][7]. In reality, change detection (CD) in SAR images is crucial in these applications, which seeks to precisely identify the changed and unchanged parts by analyzing two or more SAR images acquired over the same geographic region at different times [2,3,[7][8][9]. However, SAR images exhibit diversified inherent characteristics, such as ubiquitous multiplicative speckle noise and geometrical distortions, that inevitably impose some challenges in SAR image CD [8][9][10].…”

“…In reality, change detection (CD) in SAR images is crucial in these applications, which seeks to precisely identify the changed and unchanged parts by analyzing two or more SAR images acquired over the same geographic region at different times [2,3,[7][8][9]. However, SAR images exhibit diversified inherent characteristics, such as ubiquitous multiplicative speckle noise and geometrical distortions, that inevitably impose some challenges in SAR image CD [8][9][10].…”

“…The challenges of SAR image CD mainly come from three aspects: (1) the speckle noise and pseudochange [3,[9][10][11][12], (2) the scarcity of accessible labeled samples, and (3) class imbalance [13].…”

“…Speckle Noise and Pseudochange: Due to the particular imaging mechanism that creates SAR images by processing the radar backscatter responses coherently [10,11], speckle noise inevitably appears all over the images, which dramatically affects the intensity of the SAR images; consequently, changed pixels share a wide range of intensity values with unchanged pixels, namely intensity fluctuations. This is because speckle causes dramatic intensity fluctuations and further results in the overlapped nature between the changed and unchanged classes [9,10], which brings difficulty for accurate change feature extraction. In addition, pseudochanges can also be caused by the slight variation of the acquisition parameters, such as the imaging configuration, incidence angles, and radiometric variations, making it difficult to detect changes of interest precisely [14,15].…”