Search

Monitoring forest dynamics is of critical importance for both sustainable forest management and conservation purposes. Light detection and ranging (lidar) data provide a detailed representation of the 3-D structure of forest stands that can be used to analyze a number of trees and stand characteristics. Recently, multiple lidar acquisitions over the same area are becoming more common allowing changes in stand attributes to be assessed over time. In order to effectively utilize such multitemporal data sets for forest dynamics monitoring, we propose a method for unsupervised change detection (CD) of lidar data based on polar change vector analysis (CVA). The proposed method involves extracting relevant lidar point cloud metrics for a given area over time. Pixel-wise difference vectors of the metrics are then converted from Cartesian to polar coordinates to represent the magnitude and direction of change. Finally, the change vectors are analyzed in the polar domain to automatically discriminate between the different classes of change. The method is applied to a multitemporal lidar data set of coniferous forest on Vancouver Island, British Columbia, Canada, impacted by various types of land cover change. The experimental results demonstrate that the proposed method is capable of automatically discriminating between different classes of lidar change.

Sparse unmixing (SU) has been widely investigated for hyperspectral analysis with the aim to find the optimal subset of spectral signatures in a spectral library (known in advance) that can optimally model each pixel of the given hyperspectral image. Usually, the available spectral library organizes spectral signatures in groups. However, most existing strategies do not take full advantage of the inherent properties in the library. In this article, we design a convex framework for SU that incorporates the group structure of the spectral library. The convex framework includes two kinds of algorithms derived from either the primal or the dual form of the alternating direction method of multipliers (ADMM). Then, the convergence properties of the convex framework are established. Based on the convex framework, a novel nonconvex framework is developed for unmixing, which provides a new manner to enhance the sparsity of solution. The core of the nonconvex framework is to design a nonconvex penalty function for efficient minimization utilizing the generalized shrinkage mapping. The penalty function can be regarded as a closer approximation of the l₀ norm. Experiments conducted on simulated and real hyperspectral data demonstrate the superiority and effectiveness of the proposed nonconvex framework in improving the unmixing performance and enhancing the sparsity of solution with respect to state-of-the-art techniques.

Long-range contextual information is crucial for the semantic segmentation of High-Resolution (HR) Remote Sensing Images (RSIs). However, image cropping operations, commonly used for training neural networks, limit the perception of long-range contexts in large RSIs. To overcome this limitation, we propose a Wide-Context Network (WiCoNet) for the semantic segmentation of HR RSIs. Apart from extracting local features with a conventional CNN, the WiCoNet has an extra context branch to aggregate information from a larger image area. Moreover, we introduce a Context Transformer to embed contextual information from the context branch and selectively project it onto the local features. The Context Transformer extends the Vision Transformer, an emerging kind of neural network, to model the dual-branch semantic correlations. It overcomes the locality limitation of CNNs and enables the WiCoNet to see the bigger picture before segmenting the land-cover/land-use (LCLU) classes. Ablation studies and comparative experiments conducted on several benchmark datasets demonstrate the effectiveness of the proposed method. In addition, we present a new Beijing Land-Use (BLU) dataset. This is a large-scale HR satellite dataset with high-quality and fine-grained reference labels, which can facilitate future studies in this field.

Deep learning has attracted intensive attention in synthetic aperture radar (SAR) automatic target recognition (ATR). Usually, a considerable number of labeled samples are necessary to learn a deep model for obtaining good generalization capability. However, the process of sample labeling is time-consuming and costly. This letter proposes an active self-paced deep learning (ASPDL) approach to SAR ATR. In a nutshell, we first introduce the Bayesian inference into the process of deep model parameter optimization, aiming at learning a robust classification model in the case of a limited number of labeled samples. Next, a cost-effective sample selection strategy is presented to iteratively and actively select the informative samples from a pool of unlabeled samples for labeling. Concretely, high-confidence samples are actively selected through self-paced learning (SPL) way and automatically pseudo-labeled with the current classification model, whereas low-confidence samples are chosen through an active learning strategy and manually labeled. Finally, we update the parameters of the model by minimizing a dual-loss function using a new training set that is constructed by incorporating new labeled samples with original ones. Experiments on the moving and stationary target acquisition and recognition (MSTAR) benchmark data demonstrate that the proposed method can achieve better classification accuracy with relatively few labeled samples compared with some state-of-the-art methods.

With the increasing availability and resolution of satellite sensor data, multispectral (MS) and panchromatic (PAN) images are the most popular data that are used in remote sensing among applications. This paper proposes a novel cross-resolution hidden layer features fusion (CRHFF) approach for joint classification of multi-resolution MS and PAN images. In particular, shallow spectral and spatial features at a global scale are firstly extracted from a MS image. Then deep cross-resolution hidden layer features extracted from MS and PAN are fused from patches at a local scale according to an Autoencoder (AE) like deep network. Finally, the selected multi-resolution hidden layer features are classified in a supervised manner. By taking advantage of integrated shallow-to-deep and global-to-local features from the high-resolution MS and PAN images, the cross-resolution latent information can be extracted and fused in order to better model imaged objects from the multi-model representation, and finally increase the classification accuracy. Experimental results obtained on three real multiresolution data sets covering complex urban scenarios confirm the effectiveness of the proposed approach in terms of higher accuracy and robustness with respect to literature methods.

Deep learning has gained broad interest in remote sensing image scene classification thanks to the effectiveness of deep neural networks in extracting the semantics from complex data. However, deep networks require large amounts of training samples to obtain good generalization capabilities and are sensitive to errors in the training labels. This is a problem in remote sensing since highly reliable labels can be obtained at high costs and in limited amount. However, many sources of less reliable labeled data are available, e.g., obsolete digital maps. In order to train deep networks with larger datasets, we propose both the combination of single or multiple weak sources of labeled data with a small but reliable dataset to generate multisource labeled datasets and a novel training strategy where the reliability of each source is taken into consideration. This is done by exploiting the transition matrices describing the statistics of the errors of each source. The transition matrices are embedded into the labels and used during the training process to weigh each label according to the related source. The proposed method acts as a weighting scheme at gradient level, where each instance contributes with different weights to the optimization of different classes. The effectiveness of the proposed method is validated by experiments on different datasets. The results proved the robustness and capability of leveraging on unreliable source of labels of the proposed method.

Generative models are typically employed to approximate the distribution of deep features. Recently, these state-of-the-art methods have been applied to estimate image transformations by an unsupervised learning approach. In this letter, a novel spectral index generative adversarial network (SIGAN) is proposed for the generation of multispectral (MS) remote sensing images. This network is defined to effectively perform data augmentation starting from a limited number of training samples in the MS remote sensing domain for training deep learning models. The SIGAN model is able to capture class-specific properties in data augmentation, by incorporating the task-specific normalized spectral indices to model class-by-class properties of MS images. Experimental results obtained on a Sentinel 2 dataset show that the proposed model provides better performance than other generative adversarial networks (GANs) in MS data generation.

Deep learning-based unsupervised change detection (CD) methods compare a prechange and a postchange image in deep feature space and require precise knowledge of the event date for selecting proper pre-/post-change images. However, in many applications changes may occur gradually over a span of time making pre-/post-dates difficult to establish or prior knowledge of event date is unknown. On the other hand, deep learning-based time-series analysis methods are generally supervised. Considering such scenarios, we propose a novel unsupervised deep learning-based method to detect changes in an image time-series. The method does not make any assumption on the date of the occurrence of the change event. It treats CD as an anomaly detection problem by exploiting multilayer long short term memory (LSTM) network to learn a representation of the time series. The proposed method ingests a shuffled time series and uses an encoder-decoder LSTM model to rearrange the input sequence in correct order. While the model fails to rearrange the changed pixels, unchanged data can be rearranged in the correct order. This enables the identification of the changed pixels. To show the effectiveness of the proposed method, we tested it on two multitemporal Sentinel-1 data sets over Brumadinho, Brazil, and Bhavanisagar, India.

There are limited studies on the semantic segmentation of high-resolution polarimetric synthetic aperture radar (PolSAR) images due to the scarcity of training data and the complexity of managing speckle noise. The Gaofen contest has provided open access a high-quality PolSAR semantic segmentation dataset. Taking this opportunity, we propose a multipath residual network (MP-ResNet) architecture for the semantic segmentation of high-resolution PolSAR images. Compared to conventional U-shape encoder-decoder convolutional neural network (CNN) architectures, the MP-ResNet learns semantic context with its parallel multiscale branches, which greatly enlarges its valid receptive fields and improves the embedding of local discriminative features. In addition, MP-ResNet adopts a multilevel feature fusion design in its decoder to effectively exploit the features learned from its different branches. Comparisons with the baseline method of fully connected network (FCN with ResNet34) show that the MP-ResNet has achieved significant accuracy improvements. It also surpasses several state-of-the-art methods in terms of overall accuracy (OA), mF₁ and frequency weighted intersection over union (fwIoU), with only a limited increase of computational costs. This CNN architecture can be used as a baseline method for future studies on the semantic segmentation of PolSAR images. The code is available at: https://github.com/ggsDing/SARSeg.

Convolutional neural networks (CNNs) have achieved tremendous success in computer vision tasks, such as building extraction. However, due to domain shift, the performance of the CNNs drops sharply on unseen data from another domain, leading to poor generalization. As it is costly and time-consuming to acquire dense annotations for remote-sensing (RS) images, developing algorithms that can transfer knowledge from a labeled source domain to an unlabeled target domain is of great significance. To this end, we propose a novel full-level domain adaptation network (FDANet) for building extraction by combining image-, feature-, and output-level information effectively. At the input level, a simple Wallis filter method is employed to transfer source images into target-like ones whereby alleviating radiometric discrepancy and achieving image-level alignment. To further reduce domain shift, adversarial learning is used to enforce feature distribution consistency constraints between the source and target images. In this way, feature-level alignment can be embedded effectively. At the output level, a mean-teacher model is introduced to enforce transformation-consistent constraint for the target output so that the regularization effect is enhanced and the uncertain predictions can be suppressed as much as possible. To further improve the performance, a novel self-training strategy is also employed by using pseudo labels. The effectiveness of the proposed FDANet is verified on three diverse high-resolution aerial datasets with different resolutions and scenarios. Extensive experimental results and ablation studies demonstrated the superiority of the proposed method.

Lava tubes are buried channels that transport thermally insulated lava. Nowadays, lava tubes on the Moon are believed to be empty and thus indicated as potential habitats for humankind. In recent years, several studies investigated possible lava tube locations, considering the gravity anomaly distribution and surficial volcanic features. This article proposes a novel and unsupervised method to map candidate buried empty lava tubes in radar sounder data (radargrams) and extract their physical properties. The approach relies on a model that describes the geometrical and electromagnetic (EM) properties of lava tubes in radargrams. According to this model, reflections in radargrams are automatically detected and analyzed with a fuzzy system to identify those associated with lava tube boundaries and reject the others. The fuzzy rules consider the EM and geometrical properties of lava tubes, and thus, their appearance in radargrams. The proposed method can address the complex task of identifying candidate lava tubes on a large number of radargrams in an automatic, fast, and objective way. The final decision on candidate lava tubes should be taken in postprocessing by expert planetologists. The proposed method is tested on both a real and a simulated data set of radargrams acquired on the Moon by the Lunar Radar Sounder (LRS). Identified candidate lava tubes are processed to extract geometrical parameters, such as the depth and the thickness of the crust (roof).

The current remote sensing (RS) open data policy for multispectral (MS) missions such as Sentinel-2 and Landsat-8, together with the availability of free cloud distributed processing platforms such as Google Earth Engine, makes it possible the quick generation of burned area (BA) products even for nonexperts in the field. Indeed, fires and BAs can be detected using burn severity indices, which are usually obtained by simple band algebra operations. However, simple approaches can aid BA estimation only if typical error patterns are known and accounted for, especially when working at large (e.g., continental) scales. This article proposes an automatic BA detection system based on burn severity index thresholding, which integrates dedicated false and missed alarm mitigation strategies to improve the detection accuracy. The system is tested on Sentinel-2 and Landsat-8 data over ten different locations in Europe and spanning year 2018. Three known burn severity indices plus a custom one defined to improve the performance in the considered study area are under study. Results show that burned index thresholding is possible within accuracy bounds slightly larger than the state of the art, which is acceptable by considering the proposed simplified processing framework.

Research based on Mars Advanced Radar for Subsurface and Ionosphere Sounding (MARSIS) data detected unusual radar bright basal reflections located at about 1.5 km depth in a Mars region denoted as Ultimi Scopuli. These reflections were interpreted as a signature of subglacial liquid water even though this interpretation is still being debated in the literature. In this article, we propose a novel approach to the detection and imaging of candidate subglacial liquid water from radar sounding data. The approach combines the radar echo power traces with a suitable digital elevation model to provide a bidimensional representation of the surface. Even if the imaging method reconstructs a representation of the surface, we prove that it can be used to identify subsurface bright reflections in icy regions. Imaging is feasible even if the basal interface is not directly included in the processed data for image generation. To support this experimental evidence, we show that a relationship exists between the value of the reflected echo power originating from the englacial layers and the basal-to-surface-echo-power ratio. The observed relationship holds on both Ultimi Scopuli radar sounding data acquired on Mars and Lake Vostok data acquired on Earth. Our results show that the 2-D imaging provides an alternative way for locating candidate subglacial liquid water bodies on Mars over large areas also where the basal interface is not directly measured. The proposed approach complements previous research for further evaluation of the actual presence of liquid water on Mars.

A satellite mission onboard a radar sounder for the observation of the earth's polar regions can greatly support the monitoring of the cryosphere and climate change analyses. Several studies are in progress proposing the design and demonstrating the performance of such an earth-orbiting radar sounder (EORS). However, one critical aspect of the cryospheric targets that are often ignored and simplified in these studies is the complex geoelectrical nature of the polar ice. In this article, we present a performance assessment of the polar ice target detectability by focusing on their realistic representation. This is obtained by simulating the orbital radargrams corresponding to different regions of the polar cryosphere by leveraging the data available from airborne campaigns in Antarctica and Greenland. We propose novel performance metrics to analyze the detectability of the internal reflecting horizons (IRHs), the basal interface, and to analyze the nature of the basal interface. This performance assessment strategy can be applied to guide the design of the signal-to-noise ratio (SNR) budget at the surface, which can further support the selection of the main orbital instrument parameters, such as the transmitted power, the two-way antenna gain, and the processing gains.

Band selection (BS) can mitigate the “curse of dimensionality” problem and improve the performance of hyperspectral image (HSI) classification. Genetic algorithms (GAs) have been applied to the task of hyperspectral BS showing significant advantages compared with other literature methods. However, the traditional GAs-based methods often select sets of bands having residual redundancy due to the large search space related to hyperspectral BS and the limitation of premature convergence in GAs. Moreover, existing GAs-based methods often are supervised, and that needs a large number of labeled samples to compute the fitness value for assessing the quality of selected bands. In this article, an unsupervised BS approach based on an improved GA is proposed. A fitness function based on the fisher score combined with superpixel is designed for evaluating the discriminability of band subsets considering both spectral and spatial information. Then, modified genetic operations are constructed to restrain the search space and reduce the redundancy of selected bands. The performance of the proposed spectral-spatial GA-based BS method is evaluated on three HSIs. The experimental results demonstrate that the proposed method is superior to the traditional GA-based method and seven state-of-the-art unsupervised methods.

Convolutional neural networks (CNNs) are widely used in the field of remote sensing images. However, the applications of CNNs and related techniques often ignore the properties of remote sensing data. In our study, we focus on the hyperspectral image (HSI) classification problem, and address the issue of including the very rich spectral information present in HSIs in CNN-based models to produce highly accurate classification results. We propose a two-step classification technique, ClusterCNN. The first step divides HSI pixels into different clusters, to form a material map which can be considered as a compressed expression of the original spectral features. The second step trains a CNN that can extract spatial features from the material map, and then exploits these spatial features to classify HSI pixels. The proposed approach follows a strict hierarchy to exploit both the spectral and spatial features in HSIs. Experimental results show the effectiveness of ClusterCNN as compared to the much more complicated state-of-the-art approaches.

Hyperspectral images (HSIs) provide rich spectral–spatial information with stacked hundreds of contiguous narrowbands. Due to the existence of noise and band correlation, the selection of informative spectral–spatial kernel features poses a challenge. This is often addressed by using convolutional neural networks (CNNs) with receptive field (RF) having fixed sizes. However, these solutions cannot enable neurons to effectively adjust RF sizes and cross-channel dependencies when forward and backward propagations are used to optimize the network. In this article, we present an attention-based adaptive spectral–spatial kernel improved residual network ( A2S2K-ResNet ) with spectral attention to capture discriminative spectral–spatial features for HSI classification in an end-to-end training fashion. In particular, the proposed network learns selective 3-D convolutional kernels to jointly extract spectral–spatial features using improved 3-D ResBlocks and adopts an efficient feature recalibration (EFR) mechanism to boost the classification performance. Extensive experiments are performed on three well-known hyperspectral data sets, i.e., IP, KSC, and UP, and the proposed A2S2K-ResNet can provide better classification results in terms of overall accuracy (OA), average accuracy (AA), and Kappa compared with the existing methods investigated. The source code will be made available at https://github.com/suvojit- $0\times 55$ aa/A2S2K-ResNet.

Atmospheric phase screen (APS) is a very critical issue for the application of interferometric synthetic aperture radar (InSAR) techniques. The spatial–temporal variations of APS are the dominant error source in interferograms and may completely mask displacement signals. Many external meteorological data-based methods and phase-based methods have been developed in the past decades, but all have their inherent limitations. In this article, we propose a deep learning-based method, which is based on an attention-based deep residual U-shaped network (ARU-Net), to mitigate atmospheric artifacts. With this approach, APS patches and clean interferogram patches are sampled from InSAR interferograms to train the network. After training, the network can be used to mitigate the APS for individual interferograms. Compared with the generic atmospheric correction model (GACOS) and the advanced time-series InSAR method distributed scatterer interferometry (DSI), the key advantage of our method is that atmospheric delay can be effectively learned and removed from individual high-resolution interferometric phase itself without external data. Accuracy was validated by using individual and stacked interferograms from TerraSAR-X data over the Hong Kong International Airport (HKIA) and Hong Kong Science Park (HKSP) sites. The results showed that our method consistently delivered greater standard deviation (SD) reduction after APS correction than the GACOS method. Moreover, the time-series results were in agreement with the DSI and leveling measurements. The effectiveness of the proposed ARU-Net to remove APS effects from interferograms shows great potential for the development of a new set of deep learning-based APS reduction methods.

Change detection (CD) is one of the main applications of remote sensing. With the increasing popularity of deep learning, most recent developments of CD methods have introduced the use of deep learning techniques to increase the accuracy and automation level over traditional methods. However, when using supervised CD methods, a large amount of labeled data is needed to train deep convolutional networks with millions of parameters. These labeled data are difficult to acquire for CD tasks. To address this limitation, a novel semisupervised convolutional network for CD (SemiCDNet) is proposed based on a generative adversarial network (GAN). First, both the labeled data and unlabeled data are input into the segmentation network to produce initial predictions and entropy maps. Then, to exploit the potential of unlabeled data, two discriminators are adopted to enforce the feature distribution consistency of segmentation maps and entropy maps between the labeled and unlabeled data. During the competitive training, the generator is continuously regularized by utilizing the unlabeled information, thus improving its generalization capability. The effectiveness and reliability of our proposed method are verified on two high-resolution remote sensing data sets. Extensive experimental results demonstrate the superiority of the proposed method against other state-of-the-art approaches.

To overcome the limited capability of most state-of-the-art change detection (CD) methods in modeling spatial context of multispectral high spatial resolution (HR) images and exploiting all spectral bands jointly, this letter presents a novel unsupervised deep-learning-based CD method that can effectively model contextual information and handle the large number of bands in multispectral HR images. This is achieved by exploiting all spectral bands after grouping them into spectral-dedicated band groups. To eliminate the necessity of multitemporal training data, the proposed method exploits a data set targeted for image classification to train spectral-dedicated Auxiliary Classifier Generative Adversarial Networks (ACGANs). They are used to obtain pixelwise deep change hypervector from multitemporal images. Each feature in deep change hypervector is analyzed based on the magnitude to identify changed pixels. An ensemble decision fusion strategy is used to combine change information from different features. Experimental results on the urban, Alpine, and agricultural Sentinel-2 data sets confirm the effectiveness of the proposed method.

Radar sounder surface imaging is a rather unexplored approach to the analysis of planetary bodies. While a radar sounder is an instrument specifically designed for subsurface investigations, a particular set of power measurements (denoted as off-nadir surface echoes) can be exploited together with an external digital elevation model to produce images of the investigated surface at meters wavelength. The use of the off-nadir data may also reveal the presence of previously undetected subsurface features. In this article, we present a method for producing surface roughness images by high-frequency (HF) radar sounder data. The study of surface roughness in the HF band is particularly useful for both geologic studies and landing-zone reconnaissance as it is evaluated at meters to hundreds of meters horizontal scale. The proposed method combines off-nadir data of the Shallow Radar Sounder (SHARAD) with the Mars Orbiter Laser Altimeter (MOLA) digital elevation model. The produced roughness images at 20 MHz (15-m wavelength) of the Martian surface provide higher coverage and resolution of the surface roughness characterization at a 10–100-m horizontal scale than previous SHARAD work. By comparing the experimental roughness image with the one obtained by radar backscattering simulations, it is possible to identify subsurface features. In our experiments, we were able to produce a bidimensional image of a previously undetected large buried crater (10 km $\times12$ km) located in the Nili Fossae. This finding opens up new possibilities in exploiting radar sounding data for better detecting shallow subsurface features.

Most change detection (CD) methods are unsupervised as collecting substantial multitemporal training data is challenging. Unsupervised CD methods are driven by heuristics and lack the capability to learn from data. However, in many real-world applications, it is possible to collect a small amount of labeled data scattered across the analyzed scene. Such a few scattered labeled samples in the pool of unlabeled samples can be effectively handled by graph convolutional network (GCN) that has recently shown good performance in semisupervised single-date analysis, to improve change detection performance. Based on this, we propose a semisupervised CD method that encodes multitemporal images as a graph via multiscale parcel segmentation that effectively captures the spatial and spectral aspects of the multitemporal images. The graph is further processed through GCN to learn a multitemporal model. Information from the labeled parcels is propagated to the unlabeled ones over training iterations. By exploiting the homogeneity of the parcels, the model is used to infer the label at a pixel level. To show the effectiveness of the proposed method, we tested it on a multitemporal Very High spatial Resolution (VHR) data set acquired by Pleiades sensor over Trento, Italy.

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).

The trade-off between feature representation power and spatial localization accuracy is crucial for the dense classification/semantic segmentation of remote sensing images (RSIs). High-level features extracted from the late layers of a neural network are rich in semantic information, yet have blurred spatial details; low-level features extracted from the early layers of a network contain more pixel-level information but are isolated and noisy. It is therefore difficult to bridge the gap between high- and low-level features due to their difference in terms of physical information content and spatial distribution. In this article, we contribute to solve this problem by enhancing the feature representation in two ways. On the one hand, a patch attention module (PAM) is proposed to enhance the embedding of context information based on a patchwise calculation of local attention. On the other hand, an attention embedding module (AEM) is proposed to enrich the semantic information of low-level features by embedding local focus from high-level features. Both proposed modules are lightweight and can be applied to process the extracted features of convolutional neural networks (CNNs). Experiments show that, by integrating the proposed modules into a baseline fully convolutional network (FCN), the resulting local attention network (LANet) greatly improves the performance over the baseline and outperforms other attention-based methods on two RSI data sets.

Information from polarimetric synthetic aperture radar (PolSAR) imagery has been used for detecting built-up targets in classification problems, whereas it has been poorly exploited for change detection in multitemporal images. In this letter, we proposed an unsupervised approach for the detection of built-up changed areas from multitemporal full-polSAR images. The approach is based on the automatic thresholding of a novel change index based on the joint use of polarimetric span and average-alpha multitemporal information. The index is proposed for highlighting both constructed and demolished built-up elements. The experimental results on multitemporal UAVSAR images demonstrate that the proposed approach provides high detection accuracy and effectively separates among different types of changes, which is not the case with standard methods.