Classification-Aided SAR and AIS Data Fusion for Space-Based Maritime Surveillance

Abstract. An analysis of ice thickness distribution is a challenge, particularly in a seasonal sea ice zone with a strongly dynamic ice motion field, such as the Gulf of St. Lawrence off Canada. We present a novel automated method for ice concentration and thickness analysis combining modeling of sea ice thermodynamics and detection of ice motion on the basis of space-borne Synthetic Aperture Radar (SAR) data. Thermodynamic evolution of sea ice thickness in the Gulf of St. Lawrence was simulated for two winters, 2002–2003 and 2008–2009. The basin-scale ice thickness was controlled by atmospheric forcing, but the spatial distribution of ice thickness and concentration could not be explained by thermodynamics only. SAR data were applied to detect ice motion and ice surface structure during these two winters. The SAR analysis is based on estimation of ice motion between SAR image pairs and analysis of the local SAR texture statistics. Including SAR data analysis brought a significant added value to the results based on thermodynamics only. Our novel method combining the thermodynamic modeling and SAR yielded results that well match with the distribution of observations based on airborne Electromagnetic Induction (EM) method. Compared to the present operational method of producing ice charts for the Gulf of St. Lawrence, which is based on visual interpretation of SAR data, the new method reveals much more detailed and physically based information on spatial distribution of ice thickness. The algorithms can be run automatically, and the final products can then be used by ice analysts for operational ice service. The method is globally applicable to all seas where SAR data are available.

Ship surveillance using multiple remote sensing sensors becomes more and more vital presently. Among the various sensors, space-borne Synthetic Aperture Radar (SAR) is optimal for its high resolution over wide swaths and all-weather working capabilities. Meanwhile, Automatic Identification System (AIS) is efficient to provide ship navigational information. Limited to the progress of ship surveillance using SAR image only, the integration of them significantly benefits more. Data association is the fundamental issue. Many algorithms have been developed including the Nearest-Neighbour (NN) algorithm, the Joint Probabilistic Data Association (JPDA) method, and the Multiple Hypothesis Testing (MHT) approach. Ship positions derived from SAR image can be associated with the ones provided by AIS. State-of-the-art method (NN algorithm) is proved to be feasible. But it faces more challenges under adverse circumstances, such as high-density-shipping condition. We investigate the non-rigid Point Pattern Matching (PPM) method to solve this problem. To the best of our knowledge, this paper is the first to introduce non-rigid PPM to the data association of SAR and AIS. On the basis of introduction to the data association, Coherent Point Drift (CPD) algorithm is investigated. Experiments are carried out and the results illustrate that the CPD algorithm achieves higher accuracy and outperforms state-of-the-art method, especially under high-density-shipping condition.

ABSTRACTRecently, the performance of high-resolution spaceborne synthetic aperture radar (SAR) data for urban landscape monitoring has attracted increased interest, in particular for regions where rainy weather prevails throughout of year. In this study, the phase component of SAR data was exploited for landscape monitoring in the western area of Nanjing city, China. Twenty-six scenes of TerraSAR-X single-look complex (SLC) interferometric images were acquired for the period of May 2013 to February 2015 and applied for urban change detection as well as surface subsidence monitoring by means of interferometric SAR (InSAR) technologies. Firstly, a small baseline strategy in conjunction with a fast statistically homogeneous pixel selection (FaSHPS) approach were applied for the coherence imagery generation and for the coherent change detection, resulting in a correct probability of better than 90%. Second, the surface deformation was derived with millimetre accuracy using an enhanced multi-temporal InSAR (MT-InSAR) tool by synergistically analyzing the amplitudes and phases of the SAR acquisitions. This study indicates that high-resolution SAR data have the potential for urban landscape monitoring by taking advantage of the interferometric performance and the all-day and all-weather imaging capability.

The waterline is the transient boundary between land and sea, and the waterline extraction is of great significance for coastal research and monitoring. Spaceborne synthetic aperture radar (SAR) has the ability to penetrate clouds and thus can image large coastal areas with high-resolution under all-weather and all-day conditions. These advantages make SAR well complement the waterline observations by optical satellites. However, SAR images of coastal areas usually have complex patterns because of the diverse topography and geomorphology, and the SAR-imaging mechanism also results in large coherent speckle noise in the images. The above poses challenges to the waterline extraction from SAR imagery. Traditional manual annotation is highly accurate, but is laborious and time-consuming. Therefore, a more efficient method is urgent. The procedure of waterline extraction is sea-land segmentation followed by boundary-extraction where accurate segmentation is critical. From the perspective of the computer vision, sea-land segmentation can be viewed as the image semantic segmentation task. In recent years, deep neural network (DNN) models have achieved great success in the semantic segmentation task. The U-Net, named for its U-shape architecture, is one of the most popular DNN models for the task. Compared with others, the U-Net has a lighter structure that efficiently extract and fuse the multi-scale image features. In this study, we applied the U-Net and its variants to sea-land segmentation and systematically investigated the models’ performances. All models can segment the sea and land efficiently, and the results are overall consistent with the manual annotation. Our study demonstrates the application potential of DNN technology in waterline monitoring from SAR imagery.

Spaceborne Synthetic Aperture Radar (SAR) is well adapted to detect ocean pollution independently from daily or weather condition. In fact, oil slicks have specific impact on ocean wave spectra. Initial wave spectra may be characterized by three kinds of waves, big, medium and small, which correspond physically to gravity and gravity-capillary waves. The increase of viscosity is due to the presence of oil damps gravity-capillary waves. This induces a damping of the backscattering to the sensor, but also a damping of the energy of the wave spectra. Thus, local segmentation of wave spectra may help oil slick detection. It can be achieved by the segmentation of a multiscale decomposition of the original SAR image. In this work, a supervised oil slick detection is proposed by using Support Vector Machines into the wavelet decomposition of a SAR image. It performs accurate detection with no consideration to signal stationarity nor to the presence of strong backscatters (such as ship). Moreover, when using normalized SAR images, the kernel expansion may be generalized from one image to an other to make a near unsupervised detection scheme. The algorithm has been applied on Envisat ASAR images. First experiments yield accurate segmentation results with a very limited number of false alarms.

The extraction of ground control points (GCPs) from spaceborne synthetic aperture radar (SAR) is an effective technique for rapid, contactless, and globally available georeferencing. In recent years, China has made significant progress in the deployment and application of SAR satellites. However, the automatic derivation of high-precision GCPs remains challenging, primarily due to two key factors: 1) the need for high-accuracy calibration of SAR geometric observations, and 2) the automatic detection and matching of candidate GCPs (CGCPs) across SAR images acquired from different look angles. This paper proposed an improved algorithm framework for GCP extraction, and demonstrated the application potential of China's SAR satellites. The main improvements are as follows: 1) Beam-dependent biases are incorporated into the geometric calibration process to achieve higher ranging accuracy. 2) A CGCP detector is designed that leverages shape characteristics to robustly identify CGCPs without relying on any external data. 3) An integrated matching scheme is proposed, which combines preprocessing, feature description, and transformation-search strategies to enable effective pairing of CGCPs. SAR imagery from China's Gaofen-3B (GF3B) and Superview Neo-2 (SVN2) 01/02 satellites is utilized, and geometric calibration is performed using the open-access corner reflector array in Australia. GCP extraction experiments are conducted in the Suzhou area, with 7,477 GCPs automatically extracted and a root mean square error (RMSE) better than 1 meter.

Sensitivity of SAR data to post-fire forest regrowth in Mediterranean and boreal forests

Multitemporal single (HH) and dual-polarization (i.e., HH, HV) L-band spaceborne synthetic aperture radar (SAR) scenes were evaluated under different moisture conditions caused by precipitation prior to data acquisition at varying incidence angles. The changes affecting backscattering intensity, polarimetric decomposition, backscattering mechanism, and land use/land cover classification performance were evaluated. The study area is a shifting-cultivation environment in the eastern Amazon (Brazil). Several data input scenarios were proposed in the classification scheme (i.e., backscattering intensity alone and combined with alpha/entropy decomposition parameters, band ratios, and textural parameters) using a random forest classifier framework. Integration with optical data was also examined. The classification accuracy scores were then compared with accumulated precipitation data. The results showed that the variation in both the vegetation moisture and incidence angle increases the backscattering intensity for pasture, riparian forest and young regenerated forest by at least 1 dB compared with old successional forest stages due to its more uniform vertical structure and the landscape's increased dielectric constant. The overall classification accuracy proved low for each SAR acquisition date compared with the performance of the Landsat data. Based on SAR data, misclassification occurs for the young successional forest stages and increases in scenes with higher moisture conditions. The classification performance benefits from data integration only for one SAR scene acquired in the dry season. The results highlight the importance of selecting proper temporal intervals for the different SAR polarization modes of the forthcoming SAR missions. Further investigations should address both multitemporal at a single frequency as well as multifrequency SAR approaches.

Spaceborne Synthetic Aperture Radar (SAR) instruments are effective tools for monitoring and mapping wetlands. With the availability of SAR instruments providing various polarization options, the scope of this study is to evaluate the new compact polarization for wetland multitemporal change detection using simulated RADARSAT Constellation Mission (RCM) SAR data. A series of fully polarimetric (FP) SAR images were collected over a test site located in Ontario, Canada, and used to simulate RCM compact polarimetric (CP) data. The simulated data were evaluated for multitemporal change detection and the results were compared to those from FP SAR data. WorldView imagery and water level data were used for analysis and validation of the change detection results. The study shows potential for using the CP SAR for multitemporal change detection over three major wetland classes: shallow water, marsh, and swamp. Although FP SAR was slightly more effective in multitemporal change detection compared to CP SAR, the percentage of agreement between the change detection results of FP and CP SAR was always greater than 90% for all wetland classes. The highest overall percentage of agreement (98.4%) between the results of FP and CP SAR was observed over the shallow water, while the lowest (92.5%) was observed over swamp.

Field-scale soil moisture estimation using sentinel-1 GRD SAR data

The feasibility of measuring volcanic deformation or monitoring deformation of active volcanoes using space‐borne synthetic aperture radar (SAR) interferometry depends on the ability to maintain phase coherence over appropriate time intervals. Using ERS 1 C band (λ=5.66 cm) SAR imagery, we studied the seasonal and temporal changes of the interferometric SAR coherence for fresh lava, weathered lava, tephra with weak water reworking, tephra with strong water reworking, and fluvial deposits representing the range of typical volcanic surface materials in the Katmai volcano group, Alaska. For interferograms based on two passes with 35 days separation taken during the same summer season, we found that coherence increases after early June, reaches a peak between the middle of July and the middle of September, and finally decreases until the middle of November when coherence is completely lost for all five sites. Fresh lava has the highest coherence, followed by either weathered lava or fluvial deposits. These surfaces maintain relatively high levels of coherence for periods up to the length of the summer season. Coherence degrades more rapidly with time for surfaces covered with tephra. For images taken in different summers, only the lavas maintained coherence well enough to provide useful interferometric images, but we found only a small reduction in coherence after the first year for surfaces with lava. Measurement of volcanic deformation is possible using summer images spaced a few years apart, as long as the surface is dominated by lavas. Our studies suggest that in order to make volcanic monitoring feasible along the Aleutian arc or other regions with similar climatic conditions, observation intervals of the satellite with C band SAR should be at least every month from July through September, every week during the late spring/early summer or late fall, and every 2–3 days during the winter.

The accuracy of sequential aerial photography and SAR data for observing urban flood dynamics, a case study of the UK summer 2007 floods

Spaceborne Synthetic Aperture Radar (SAR) and Optical sensors, are one of the main sources of Earth observation in the present age. Both the data types have their inherent advantages and disadvantages. Spaceborne Optical sensor are restricted by clouds but can offer strong information content in ideal conditions. On the other hand, SAR sensors rely on their own energy and can see through clouds. SAR is potentially an all-weather day/night imager. But SAR sensors have limitations in terms of data collection geometry and algorithmic approximations. Both sensors offer complimentary information for exploitation in data fusion for enhanced results. This research is focused on capitalizing the fusion potential for spaceborne High resolution SAR and Optical data in urban settings. The fusion of high reflection of SAR energy from urban areas and optical features of such areas can be combined to enhance the urban infrastructure detection and monitoring in a SAR/Optical fused scenario. SAR/Optical fusion can take place at three levels 1) pixel level, 2) feature level; and 3) information level. Pixel level fusion is often considered most difficult for high resolution data as precise registration up to subpixel level is required and even slight misregistration results in unfavorable circumstances. Simon Fraser University (SFU) Burnaby Mountain Campus has been chosen for area of interest because of its ongoing student housing and university infrastructure developmental projects. TerraSAR-X High Resolution Spotlight (TSX-HS) Single Look Complex (SLC) images of 1.0 m resolution have continuously being acquired over SFU; along with high resolution Optical (RGB) and Infrared (IR) images (3.0 m resolution each) from “The Planet” acquisitions. Limited high-resolution images from “Google Earth” (GE) in the coinciding period of TSX-HS acquisitions were also acquired for the study. Six fusion techniques have been studied for urban infrastructure detection and have been categorized based on their performance. Precision change maps will be created based on time series analysis for SAR/optical fused data in conjunction with Interferometric SAR (InSAR) analysis to study the long-term effect of urban infrastructure developments over a period of two years.

Spaceborne Synthetic Aperture Radar (SAR) can be applied for monitoring tropical cyclones (TCs), but co-polarized C-band SAR suffers from signal saturation such that it is improper for high wind-speed conditions. In contrast, cross-polarized SAR data does not suffer from this issue, but the retrieval algorithm needs more deliberation. Previously, many geophysical model functions (GMFs) have been developed using cross-polarized data, which obtain wind speeds using the complex relationships described by radar backscatter, incidence angle, wind direction, and radar look direction. In this regard, the rapid development of artificial intelligence technology has provided versatile machine learning methods for such a nonlinear inversion problem. In this study, we comprehensively compare the wind-speed retrieval performance of several models including Back Propagation Neural Network (BPNN), Support Vector Machine (SVM), Random Forest (RF), and Deep Neural Network (DNN), which were developed based on spatio-temporal matching and correlation analysis of stepped frequency microwave radiometer (SFMR) and dual-polarized Sentinel-1 SAR data after noise removal. A data set with ~2800 samples is generated during TCs for training and validating the inversion model. The generalization ability of different models is tested by the reserved independent data. When using similar parameters with GMFs, RF inversion has the highest accuracy with a Root Mean Square Error (RMSE) of 3.40 m/s and correlation coefficient of 0.94. Furthermore, considering that the sea surface temperature is a crucial factor for generating TCs and influencing ocean backscattering, its effects on the proposed RF model are also explored, the results of which show improved wind-speed retrieval performances.

With the increase of the amount of space-borne synthetic aperture radar (SAR), SAR data is increasingly being applied for remote monitoring of hurricane wind speed filed. There are many geophysical model functions (GMFs) have been developed for wind speed inversion by describing the relationship between the surface winds and the normalized radar backscatter cross section (NRCS) of SAR. In this paper, we provide a method based on support vector machine (SVM) for retrieving oceanic surface wind speeds over hurricanes. But unlike most traditional GMFs, this method does not need formula fitting for the input parameters which include dual polarization (VV+VH) SAR normalized radar cross section and incidence angle (θ). In addition, we use the Stepped Frequency Microwave Radiometer (SFMR) wind filed data as the training and validation data for SVM model. The results show that the SVM model has achieved good results over totally independent testing data. The retrieved hurricane wind speed are in good agreement with the SFMR wind filed, and the correlation coefficient and root mean square error(RMSE) were 0.91 and 4.20 m/s, respectively.