Synthetic Aperture Radar Change Detection Research Articles

False Alarm Reduction in Wavelength–Resolution SAR Change Detection Schemes by Using a Convolutional Neural Network

In this letter, we propose a method to reduce the number of false alarms in a wavelength–resolution synthetic aperture radar (SAR) change detection scheme by using a convolutional neural network (CNN). The detection is performed in two steps: change analysis and object classification. A simple technique for wavelength–resolution SAR change detection is implemented to extract potential targets from the image of interest. A CNN is then used for classifying the change map detections as either a target or nontarget, further reducing the false alarm rate (FAR). The scheme is tested for the CARABAS-II data set, where only three false alarms over a testing area of 96 km <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> are reported while still sustaining a probability of detection above 96%. We also show that the network can still reduce the FAR even when the flight heading of the SAR system measurement campaign differs by up to 100° between the images used for training and test.

Joint exploitation of spaceborne SAR images and GIS techniques for urban coherent change detection

This paper proposes a simple and fast method for the identification of structural changes affecting buildings in urban environments by using a combination of Synthetic Aperture Radar (SAR) imagery and Geospatial Information System (GIS) processing.The identification of changes in urban settlements is of great interest for damage assessment after natural disasters, cadastral mapping and monitoring urban development or illegal activities, such as the construction of unauthorized buildings.Satellite remote sensing is useful in this scenario and SAR data is attractive due to its wide and ubiquitous coverage, the day and night all-weather availability, the exact repetition of the acquisition geometry, the repeated illumination and the sensitivity to slight changes in the geometrical structure of the targets in the scene.This sensibility is an advantage, but turns into a drawback especially in an urban environment where every subtle change may cause an unwanted detection.This environment is indeed one of the most challenging for the detection of those changes that are of any real interest since these events are masked by thousands of irrelevant detections.This paper tackle this problem with a combination of an improved, high-resolution coherent change detection technique called M-CCD and with a GIS post-processing. The result is a map of changes affecting buildings that are of a significant scale and consequently of a certain interest in an urban environment.In this contribution, the complete workflow is detailed and an assessment of the detected changes is done with high resolution optical images through visual photo-interpretation. A comparison with other SAR and optical change detection methods is also carried out.

Building Change Detection in VHR SAR Images via Unsupervised Deep Transcoding

Building change detection (CD), important for its application in urban monitoring, can be performed in near real time by comparing prechange and postchange very-high-spatial-resolution (VHR) synthetic-aperture-radar (SAR) images. However, multitemporal VHR SAR images are complex as they show high spatial correlation, prone to shadows, and show an inhomogeneous signature. Spatial context needs to be taken into account to effectively detect a change in such images. Recently, convolutional-neural-network (CNN)-based transfer learning techniques have shown strong performance for CD in VHR multispectral images. However, its direct use for SAR CD is impeded by the absence of labeled SAR data and, thus, pretrained networks. To overcome this, we exploit the availability of paired unlabeled SAR and optical images to train for the suboptimal task of transcoding SAR images into optical images using a cycle-consistent generative adversarial network (CycleGAN). The CycleGAN consists of two generator networks: one for transcoding SAR images into the optical image domain and the other for projecting optical images into the SAR image domain. After unsupervised training, the generator transcoding SAR images into optical ones is used as a bitemporal deep feature extractor to extract optical-like features from bitemporal SAR images. Thus, deep change vector analysis (DCVA) and fuzzy rules can be applied to identify changed buildings (new/destroyed). We validate our method on two data sets made up of pairs of bitemporal VHR SAR images on the city of L’Aquila (Italy) and Trento (Italy).

Near Real-Time Freeze Detection over Agricultural Plots Using Sentinel-1 Data

Short-term freeze/thaw cycles, which mostly occur in the northern hemisphere across the majority of land surfaces, are reported to cause severe economic losses over broad areas of Europe and North America. Therefore, in order to assess the extent of frost damage in the agricultural sector, the objective of this study is to build an operational approach capable of detecting frozen plots at the plot scale in a near real-time scenario using Sentinel-1 (S1) data. C-band synthetic aperture radar (SAR) data show high potential for the detection of freeze/thaw surface states due to the significant alterations to the dielectric properties of the soil, which are distinctly observable in the backscattered signal. In this study, we propose an approach that relies on change detection in the high-resolution Sentinel-1 C-band SAR backscattered coefficients, to determine surface states at the plot scale as either frozen or unfrozen. A threshold analysis is first performed in order to determine the best thresholds for three distinct land cover classes, and for each polarization mode (VH, and VV). S-1 SAR data are then used to detect a plot’s surface state as either unfrozen, mild-to-moderately frozen or severely frozen. A temperature-based filter has also been applied at the end of the detection chain to eliminate false detections in the freezing detection algorithm due mainly to rainfall, irrigation, tillage, or signal noise. Our approach has been tested over two study sites in France, and the output results, using either VH or VV, compared qualitatively well with both in situ air temperature data and soil temperature data provided by ERA5-Land. Overall, our algorithm was able to detect all freezing episodes over the analyzed S-1 SAR time series, and with no false detections. Moreover, given the high-resolution aspect of S-1 SAR data, our algorithm is capable of mapping the local variation of freezing episodes at plot scale. This is in contrast with previous products that only offer coarser results across larger areas (low spatial resolution).

Wavelength-Resolution SAR Change Detection Using Bayes’ Theorem

This article presents Bayes’ theorem for wavelength-resolution synthetic aperture radar (SAR) change detection method development. Different change detection methods can be derived using Bayes’ theorem in combination with the target model, clutter-plus-noise model, iterative implementation, and noniterative implementation. As an example of the Bayes’ theorem use for wavelength-resolution SAR change detection method development, we propose a simple change detection method with a clutter-plus-noise model and noniterative implementation. In spite of simplicity, the proposed method provides a very competitive performance in terms of probability of detection and false alarm rate. The best result was a probability of detection of $\text{98.7}\%$ versus a false alarm rate of one per square kilometer.

Change Detection From Synthetic Aperture Radar Images Based on Channel Weighting-Based Deep Cascade Network

Deep learning methods have recently demonstrated their significant capability for synthetic aperture radar (SAR) image change detection. However, with the increase of network depth, convolutional neural networks often encounter some negative effects, such as overfitting and exploding gradients. In addition, the existing deep networks employed in SAR change detection tend to produce a lot of redundant features that affect the performance of the network. To solve the aforementioned problems, this article proposed a deep cascade network (DCNet) for SAR image change detection. On the one hand, a very DCNet is established to exploit discriminative features, and residual learning is introduced to solve the exploding gradients problem. In addition, a fusion mechanism is employed to combine the outputs of different hierarchical layers to further alleviate the exploding gradient problem. Moreover, a simple yet effective channel weighting-based module is designed for SAR change detection. Average pooling and max pooling are used to aggregate channel-wise information. Meaningful channel-wise features are emphasized and unnecessary ones are suppressed. Therefore, the similarity in feature maps can be reduced, and then, the classification performance of the DCNet is improved. Experimental results on four real SAR datasets demonstrated that the proposed DCNet can obtain better change detection performance than several competitive methods. Our codes are available at https://github.com/summitgao/SAR_CD_DCNet .

Stacked Fisher autoencoder for SAR change detection

Stacked autoencoder is effective in image denoising and classification when it is used for synthetic aperture radar (SAR) change detection. However, the resulting features may not be discriminative enough for in some sense. To alleviate this problem, in this paper we propose a stacked Fisher autoencoder (SFAE) for SAR change detection. Specifically, in the framework of SFAE, unsupervised layer-wise feature learning and supervised fine-tuning are jointly performed when training the network. The trained network can be used to detect the changes in both of the single and multi-polarization SAR datasets in real-time. The proposed SFAE has two advantages. The first one is to expand the stacked autoencoder to suit the environment with the multiplicative noise in SAR change detection. The second is that the features extracted by SFAE are more discriminative than the original stacked autoencoder due to that Fisher discriminant criterion is incorporated into SFAE. The results on the simulated and real SAR datasets indicate that the proposed SFAE algorithm has a significant advantage on multitemporal single/multi-polarization SAR (SAR/PolSAR) change detection.

Unbiased Seamless SAR Image Change Detection Based on Normalized Compression Distance

Land cover changes may have very different nature, e.g., vegetation development, soil erosion, variation of humidity, or damage of buildings, only to enumerate few cases. In addition, synthetic aperture radar (SAR) observations are a doppelganger of the scene, imaging the scene signature rather than the scene itself. To overcome these challenges, SAR change detection methods generally adapt to the particular situations. We present seamless methods based on normalized compression distance (NCD) estimation. NCD is a similarity metric applied directly to the data, thus with no biases induced by feature estimators or classifiers. Since the diversity of changes is huge and extremely hard to derive typical classes, we introduce paradigm based both on an unsupervised and a supervised method. The change detection procedure mainly consists in dividing image dataset in patches, computing a collection of similarities for pairs of tiles formed differently in each case, and usage of this collection in unsupervised and supervised forms to generate a change map. Both the threshold based histogram, unsupervised method, and the k-NN classifier algorithm, supervised method, have a distinct flow to obtain the change map. To use the NCD operator according to our proposed methods, a speckle resistance test is involved. The experimental results for the two methodologies are computed using two TerraSAR-X images over Sendai and surrounding areas, Japan.

An Unsupervised SAR Change Detection Method Based on Stochastic Subspace Ensemble Learning

As synthetic aperture radar (SAR) is playing an increasingly important role in Earth observations, many new methods and technologies have been proposed for change detection using multi-temporal SAR images. Especially with the development of deep learning, numerous methods have been proposed in recent years. However, the requirement to have a certain number of high-quality samples has become one of the main reasons for the limited development of these methods. Thus, in this paper, we propose an unsupervised SAR change detection method that is based on stochastic subspace ensemble learning. The proposed method consists of two stages: The first stage involves the automatic determination of high-confidence samples, which includes a fusion strategy and a refinement process; and the second stage entails using the stochastic subspace ensemble learning module, which contains three steps: obtaining the subsample sets, establishing and training a two-channel network, and applying the prediction results and an ensemble strategy. The subsample sets are used to solve the problem of imbalanced samples. The two-channel networks are used to extract high-dimensional features and learn the relationship between the neighborhood of the pixels in the original images and the labels. Finally, by using an ensemble strategy, the results predicted by all patches reclassified in each network are integrated as the detection result. The experimental results of different SAR datasets prove the effectiveness and the feasibility of the proposed method.

SAR image change detection based on deep denoising and CNN

The intrinsic noise of synthetic aperture radar (SAR) images has a big influence to the image processing performance, especially in change detection (CD). Image denoising is an important branch of image restoration which aims at enhancing the quality of images. The detection accuracy of CD depends greatly on the quality of red difference image (DI), therefore image denoising can be regarded as a vital step in SAR CD. However, few researches focused on this problem. In this study, an end-to-end deep denoising model is first designed to remove the noise of SAR images. With the help of abundant simulated SAR images, deep denoising model is trained effectively to estimate the noise component. Then clean image can be achieved by removing this noise component from the original SAR image. After denoising, the new image pair will generate a clean DI. At last, DI is classified into changed and unchanged areas by a three-layer Convolutional Neural Network (CNN). Three real SAR image pairs demonstrate the effectiveness of the proposed method.

A SAR change detection method based on the consistency of single-pixel difference and neighbourhood difference

ABSTRACTThis letter presents a synthetic aperture radar (SAR) change detection method based on the consistency of single-pixel difference and neighbourhood difference. The single-pixel difference is reflected by the log-ratio (LR), and the logarithmic likelihood ratio (LLR) is used to calculate the neighbourhood difference. Then, the two operators mentioned are combined to construct a new detector. The new detector can enhance the degree of difference between the changed class and unchanged class while having better anti-noise ability. According to the histogram of the difference image (DI), the ratio curve of the histogram presents a pattern that the curve changes from stabilization to random oscillation. Then, the transitional point is selected as the threshold. The experimental results from different data sets demonstrate the effectiveness of the proposed method.

High-Resolution SAR Change Detection Based on ROI and SPP Net

With the resolution increasing, the structure information becomes more and more abundant in Synthetic Aperture Radar (SAR) images. The speckle noise generated by the coherent imaging mechanism, has a great influence on the detection accuracy and detection difficulty accordingly in high-resolution SAR change detection. In this paper, a multivariate change detection framework based on non-subsampled contourlet transform (NSCT), deep belief networks (DBN), fuzzy c-means (FCM) clustering, and global-local spatial pyramid pooling (SPP) net is proposed. NSCT decomposes the image into multiple scales and DBN is used for extracting feature of the decomposed coefficient matrix. FCM converges the similarity matrix of the initial features by DBN into two classes as a pseudo-label for global-local SPP net training data. The global-local SPP net consists of a large-scale region of interest (ROI) SPP net and a small-scale change detection SPP net. The combination of ROI and the SPP net, as well as the fusion between different scales, weakens the interference of the unchanged information and effectively eliminates a large number of redundant information. The experimental results show that our proposed method can effectively remove speckle noise and improve the robustness of high-resolution SAR change detection.

Bivariate Gamma Distribution for Wavelength-Resolution SAR Change Detection

A gamma probability density function (pdf) is shown to be an alternative to model the distribution of the magnitudes of high-resolution, i.e., wavelength-resolution, synthetic aperture radar (SAR) images. As investigated in this paper, it is more appropriate and more realistic statistical in comparison with, e.g., Rayleigh. A bivariate gamma pdf is considered for developing a statistical hypothesis test for wavelength-resolution incoherent SAR change detection. The practical issues in implementation of statistical hypothesis test, such as assumptions on target magnitudes, estimations for scale and shape parameters, and implementation of modified Bessel function, are addressed. This paper also proposes a simple processing scheme for incoherent change detection to validate the proposed statistical hypothesis test. The proposal was experimented with 24 CARABAS data sets. With an average detection probability of 96%, the false alarm rate is only 0.47 per square kilometer.

A SAR INTENSITY IMAGES CHANGE DETECTION METHOD BASED ON FUSION DIFFERENCE DETECTOR AND STATISTICAL PROPERTIES

Abstract. Detecting the land cover changes is an important application of multi-temporal synthetic aperture radar (SAR) images. This study puts forward a novel SAR change detection method which has two-steps: change detector construction and change threshold selection. For change detector construction, considering the SAR intensity images follow the gamma distribution, the conditional probabilities of the binary hypothesis test are provided, then the log likelihood ratio (LLR) combined with the log ratio (LR) to construct a detector which can enhance the degree of change to calculate the diversity degree convenient between the two images; for change threshold selection, owing to the characteristic that the curve about the ratio value of adjacent grey-level (GL) values in normalized difference map, the normalized difference map can be segmented in three parts by two thresholds selected which correspond to the regions of unchanged, backscatter enhanced and weakened separately. And as this, the change areas can be also determined simultaneously. The experimental results on different areas and sensors indicate that the proposed algorithm is effective and feasible.

On the Statistics of Coherence Estimators for Textured Clutter Plus Noise

This letter presents a theoretical analysis of the impact of thermal noise on the statistics of the classical as well as a modified estimator for the cross-correlation coefficient between two receive channels, which follow a noisy compound clutter model. This correlation coefficient, in the synthetic aperture radar (SAR) context often called coherence, is widely used as an important quality parameter in the field of SAR interferometry, SAR change detection, and SAR ground moving target indication. Based on a novel closed-form expressions for the probability density function (pdf) for integer number of averaged samples (or looks), it is shown that, contrary to widespread belief, the clutter texture does not cancel out when the noise contribution is taken into account. It is further demonstrated that the new pdfs can be used to derive the bias and variance of the sample coherence in analytical closed form. Thereby, the impact of the texture on the estimators is analyzed.

Learning Change from Synthetic Aperture Radar Images: Performance Evaluation of a Support Vector Machine to Detect Earthquake and Tsunami-Induced Changes

This study evaluates the performance of a Support Vector Machine (SVM) classifier to learn and detect changes in single- and multi-temporal X- and L-band Synthetic Aperture Radar (SAR) images under varying conditions. The purpose is to provide guidance on how to train a powerful learning machine for change detection in SAR images and to contribute to a better understanding of potentials and limitations of supervised change detection approaches. This becomes particularly important on the background of a rapidly growing demand for SAR change detection to support rapid situation awareness in case of natural disasters. The application environment of this study thus focuses on detecting changes caused by the 2011 Tohoku earthquake and tsunami disaster, where single polarized TerraSAR-X and ALOS PALSAR intensity images are used as input. An unprecedented reference dataset of more than 18,000 buildings that have been visually inspected by local authorities for damages after the disaster forms a solid statistical population for the performance experiments. Several critical choices commonly made during the training stage of a learning machine are being assessed for their influence on the change detection performance, including sampling approach, location and number of training samples, classification scheme, change feature space and the acquisition dates of the satellite images. Furthermore, the proposed machine learning approach is compared with the widely used change image thresholding. The study concludes that a well-trained and tuned SVM can provide highly accurate change detections that outperform change image thresholding. While good performance is achieved in the binary change detection case, a distinction between multiple change classes in terms of damage grades leads to poor performance in the tested experimental setting. The major drawback of a machine learning approach is related to the high costs of training. The outcomes of this study, however, indicate that given dynamic parameter tuning, feature selection and an appropriate sampling approach, already small training samples (100 samples per class) are sufficient to produce high change detection rates. Moreover, the experiments show a good generalization ability of SVM which allows transfer and reuse of trained learning machines.

Forcing Scale Invariance in Multipolarization SAR Change Detection

This paper considers the problem of coherent (in the sense that both amplitudes and relative phases of the polarimetric returns are used to construct the decision statistic) multipolarization synthetic aperture radar change detection starting from the availability of image pairs exhibiting possible power mismatches/miscalibrations. The principle of invariance is used to characterize the class of scale-invariant decision rules which are insensitive to power mismatches and ensure the constant false alarm rate property. A maximal invariant statistic is derived together with the induced maximal invariant in the parameter space which significantly compresses the data/parameter domain. A generalized likelihood ratio test is synthesized both for the cases of two- and three-polarimetric channels. Interestingly, for the two-channel case, it is based on the comparison of the condition number of a data-dependent matrix with a suitable threshold. Some additional invariant decision rules are also proposed. The performance of the considered scale-invariant structures is compared to those from two noninvariant counterparts using both simulated and real radar data. The results highlight the robustness of the proposed method and the performance tradeoff involved.

Two-Stage Change Detection for Synthetic Aperture Radar

Coherent change detection using paired synthetic aperture radar (SAR) images is often performed using a classical coherence estimator that is invariant to the true variances of the populations underlying each paired sample. While attractive, this estimator is biased and requires a significant number of samples to yield good performance. Increasing sample size often results in decreased image resolution. Thus, we propose the use of Berger's coherence estimate because, with the same number of pixels, the estimator effectively doubles the sample support without sacrificing resolution when the underlying population variances are equal or near equal. A potential drawback of this approach is that it is not invariant since its distribution depends on the pixel pair population variances. While Berger's estimator is inherently sensitive to the inequality of population variances, we propose a method of insulating the detector from this acuity. A two-stage change statistic is introduced to combine a noncoherent intensity change statistic given by the sample variance ratio, followed by the alternative Berger estimator, which assumes equal population variances. The first-stage detector identifies pixel pairs that have nonequal variances as changes caused by the displacement of sizeable object. The pixel pairs that are identified to have equal or near-equal variances in the first stage are used as an input to the second stage. The second-stage test uses the alternative Berger coherence estimator to detect subtle changes such as tire tracks and footprints. We show experimentally that the proposed method yields higher contrast SAR change detection images than the classical coherent change detector (state of the art), the alternative coherent change detector, and the intensity change detector. Experimental results are presented to show the effectiveness and robustness of the proposed algorithm for SAR change detection.

Manual detection of snow avalanche debris using high-resolution Radarsat-2 SAR images

Snow avalanches cause significant loss of life and infrastructure in snow covered mountain areas worldwide. Our understanding and ability to forecast avalanche activity in space and time, are limited by incomplete avalanche activity datasets, derived from point-scale field observations. Here we propose the use of C-band synthetic aperture radar (SAR) in detecting avalanche debris. We use high-resolution Radarsat-2 Ultrafine (RS-2U) imagery to manually detect avalanche debris-like features based on the backscatter contrast between avalanche debris and the surrounding, undisturbed snowpack. During March 2014, we acquired 12 RS-2U images, covering roughly 12% of the county of Troms in Northern Norway. Using single backscatter images, we detected 546 avalanche debris-like features of which 57 overlapped in multiple times, 7 were either radar shadows and layovers or natural features, and 15 were eliminated by a topographic GIS model, distinguishing avalanche from non-avalanche terrain. The remaining 467 features were classified as avalanche debris, of which 173 features (37%) could be validated. Validation was carried out in the field, by using optical remote sensing data from Landsat-8, by compiling SAR change detection images and by independent manual expert detection. SAR change detection utilizes the relative backscatter increase of avalanche debris between a snow-free SAR image and a current SAR image with avalanche activity. This relative backscatter increase was quantified with a median relative backscatter of 3.7dB, and a threshold of 1.5dB. In this study, we showed that detection of small sized avalanche debris is possible using RS-2U images. However, the operational usage is limited by the high acquisition costs, small ground swath, and uncertain data availability.

A Generalized Likelihood Ratio Test for Coherent Change Detection in Polarimetric SAR

Several detection statistics have been proposed for detecting fine ground disturbances between two synthetic aperture radar (SAR) images, such as vehicle tracks. The standard method involves estimating a local correlation coefficient between images. Other methods have been proposed using various statistical hypothesis tests. One of these alternative methods is a generalized likelihood ratio test (GLRT), which compares a full-correlation image model to a no-correlation image model. In this letter, we expand the GLRT to polarimetric SAR data and derive the appropriate GLRT detection statistics. Additionally, we explore relaxing the equal variance/equal polarimetric covariance assumptions used in previous results and find improved performance on macroscopic scene changes.