Maritime Domain Awareness (MDA) confronts significant challenges in the maritime domain, leveraging satellite technologies that play a role in enabling extensive and consistent area mapping. In this case, Synthetic Aperture Radar (SAR) stands out for its all-weather capability, serving as a crucial tool for applications ranging from environmental monitoring to defense systems (Ulaby and Long, 2014). Moreover, satellite constellations offer numerous advantages, including global coverage, regular revisit intervals, remote accessibility, and substantial data volume (Sandau et al, 2010)(Graziano et al., 2012).

Currently, Maritime Traffic Monitoring (MTM) depends on land-based systems like coastal radars and the Automatic Identification System (AIS), as well as the efforts of coast guards and port authorities. AIS is a collaborative system wherein ships equipped with transmitters share identification, position, and course data with shore stations. It facilitates ship-to-ship communication to prevent collisions (Soldi et al., 2020). The system can be deactivated by shipboard personnel under very few exceptions and its range of reception is variable, dependent on factors such as signal propagation conditions, sea state, the height of the transmitting and receiving antenna, and the strength of the vessel transmitter. On average, a 40 nautical mile reception radius is likely to be achieved by the AIS receiver network. Further, it is mandatory for all ships of 300 gross tonnage and above engaged on international voyages, and all passenger ships irrespective of size, according to the International Convention for the Safety of Life at Sea (SOLAS) Regulation V/19.2.4 (NATO Shipping Centre, 2021). An alternative with similar characteristics must be adopted, such as gathering information from various sources is crucial for the effective management of non-cooperative ship activities. Satellite images, with their unique vantage point from space, offer a promising avenue to supplement existing information sources and enhance maritime surveillance.

For instance, an illustrative study in Ghana (Kurekin et al., 2019) employed SAR satellite data to monitor uncooperative ships and combat extensive illegal fishing. The results indicated that over 75% of detected vessels were uncooperative, and Figure 1 portrays all surveyed vessels from July 2016 to December 2017. More than three-quarters of the ships detected by Sentinel-1 lacked AIS associations, implying operation without AIS or with AIS disabled.

In the upcoming sections of this article, three essential aspects are explored: firstly, the process of collecting satellite images, followed by a dive into how ship detection generally works, and subsequently introducing how SAR images are enhanced for improving maritime awareness.

Figure 1: Ship detected in Ghana from July 2016 to December 2017 (Kurekin, 2019).
Figure 1: Ship detected in Ghana from July 2016 to December 2017 (Kurekin, 2019)


How satellite images are collected: The case of Sentinel-1

The investigation of Sentinel-1 satellite operations is divided into four distinctive methodologies (Figure 2) employed to capture images through the utilization of SAR (ESA, s.d.). These diverse imaging techniques offer varying data types. They facilitate comprehension of phenomena like ocean wind patterns and wave characteristics. In essence, the process is akin to the assembly of puzzle pieces, which upon integration, yield a comprehensive portrayal of Earth's activities.

The first methodology, designated as the Wave mode (WV), acquires images that encompass approximately 20 by 20 kilometers of terrain, resembling snapshots taken from space as the satellite orbits around the Earth. Secondly, the StripMap (SM) mode yields highly detailed images, unveiling features as small as 5 by 5 meters. It is analogous to possessing an exceptionally clear camera located in space.

Additionally, the Interferometric Wide-Swath (IW) mode functions as a puzzle assembler by capturing images at slightly varied intervals. This capability allows for the identification of alterations on the Earth's surface, such as subtle movements or shifts. Lastly, the Extra-Wide-Swath (EW) mode offers a broader perspective, covering a wider expanse, although the finer details might not be as distinct.


Figure 2: Sentinel-1 modes (ESA, s.d.).
Figure 2: Sentinel-1 modes (ESA, s.d.).


Furthermore, the Sentinel-1 satellite provides an assortment of data product levels, classified as Level 0, Level 1 SLC, Level 1 GRD, and Level 2 OCN. Each level corresponds to a distinct phase of data processing, contributing to an enhanced understanding of Earth's dynamics. At Level 0, raw and unprocessed data forms the basis for subsequent data refinement. Progressing to Level 1, the data is made more user-friendly, and available in two formats: Single Look Complex (SLC) and Ground Range Detected (GRD). The former undergoes meticulous adjustment using supplementary satellite data, while the latter undergoes enhanced processing to enhance its utility. Level 1 GRD products are available in varying sizes, offering adaptability to diverse needs. Ultimately, at Level 2, the insights gleaned from earlier stages are transformed into even more valuable information, including details about ocean wind patterns and wave characteristics.

Detecting ships in satellite images

Methods that involve preliminary screening and discrimination, using data from various missions, for example, can be employed to identify ships, also those that are uncooperative but visible in the captured SAR images.

In general, algorithms are used to identify objects in images. However, it's important to note that some of these may produce a percentage of false positives (also known as false alarms), indicating cases where a portion of the sea is incorrectly identified as a ship, as well as false negatives (or missed alarms), indicating that a ship is incorrectly identified as part of the sea. At this point, it is noteworthy to mention that if an algorithm is adjusted to reduce false negatives, it might lead to an increase in false positives and vice versa. Accepting a certain number of false alarms can lead to a decrease in instances where real ships are missed in the identification process (Crisp, 2004). Figure 3 illustrates this relationship between false positives and false negatives, making it easier to understand the trade-off between the two outcomes.

Figure 3: Examples of true positive (upper left), false negative (upper right), false positive (lower left), false positive near the coast (lower right)
Figure 3: Examples of true positive (upper left), false negative (upper right), false positive (lower left), false positive near the coast (lower right)


If the goal is to minimize the percentage of false negatives, a threshold approach can be adopted (Paes et al., 2010) (Lorenzzetti et al., 2010) (Crisp, 2004). This involves identifying a ship by the cluster of bright pixels it forms against a darker background. By setting a specific threshold for pixel intensity, it is possible to distinguish ship pixels from sea pixels. The choice of this threshold leads to two different techniques: a uniform threshold for the entire image or an adaptive threshold that adjusts within the image.

Within the latter technique, the Constant False Alarm Rate (CFAR) is often used to maintain a consistent rate of false positives (Crisp, 2004). At hand, the image is scanned with three distinct “window” frames: a background window, a guard window, and a target window (Figure 4). By comparing the brightness of pixels within the target window to the ones between the background and guard windows, it is possible to detect the presence of a bright object like a ship. To achieve this, the average pixel brightness within the target window is calculated for each position along the image. This average is then compared to a threshold value derived from statistical analysis of pixels in the outer frame, ensuring that the occurrence of false positives remains steady across the process.

Figure 4: Typical window setup for an adaptive threshold detector
Figure 4: Typical window setup for an adaptive threshold detector


SAR image analysis with the Sentinel Application Platform (SNAP)

Handling data products involves the visualization and analysis of data. To accomplish this, the open-source SNAP (ESA, s.d.) can be used.

As an example of its usage, the SAR image depicted in Figure 5 (IW, Level 1 SLC) will be assessed. It is important to note that the view is flipped, as the scene was acquired during an ascending passage, meaning that the satellite was following a path in a south-north direction. Therefore, the pixels result in the order of data acquisition. Furthermore, if necessary, a procedure can be performed when the analysis needs to concentrate on a relatively compact area. This allows for avoiding the need to process the entire image, resulting in reduced processing time.

Figure 5: Sentinel-1 SAR product visualized on SNAP
Figure 5: Sentinel-1 SAR product visualized on SNAP


With the IW mode, three sub-swaths are acquired using the Terrain Observation with Progressive Scan SAR (TOPSAR) technique. Using the TOPSAR method involves not only adjusting the beam in range, similar to ScanSAR, but also electronically steering the beam from backward to forward in the azimuth direction for each burst, i.e. the horizontal “line”, preventing scalloping and ensures consistent image quality across the entire swath (De Zan and Monti Guarnieri, 2006). Using SNAP to de-burst, it is possible to have the merging of the bursts, and then the sub-swaths, to generate the complete product. This process produces the result shown in Figure 6.

Figure 6: De-burst result
Figure 6: De-burst result


At this point, the de-burst product in input is used to detect objects in the marine area of interest. SNAP can both mask the ground areas, which are not of interest and extend the coastline to reduce the number of false alarms in its vicinity. In addition, it is possible to perform changes in the CFAR algorithm’s window frame size, filtering out also the false targets based on the choice of minimum and maximum ship size.

To better refine ship detection, a performance analysis should be also carried out to minimize the number of missed detections (e.g., in the CFAR algorithm), accepting a non-negligible share of false alarms. Figure 7 shows the result of ship detection, as an example.

By subsequently exporting the results from SNAP, they can be further processed and visualized in QGIS or Google Earth (Figure 8).

Figure 7: Ship detection results
Figure 7: Ship detection results
Figure 8: QGIS visualization
Figure 8: QGIS visualization



Maritime traffic is growing rapidly, and the presence of more ships inevitably increases the chances of infractions but also of environmentally devastating ship accidents. By taking advantage of space missions, it is therefore possible to compare historical information and habitual behavior, enabling the identification of anomalous conduct and potential security threats. The result is timely and easy-to-access information that is useful for identifying possible hazards, planning actions by the relevant authorities, intercepting responsible vessels, and plotting safe routes in hostile environments. Therefore, satellite data can support the maritime industry in increasing knowledge, anticipating threats, triggering alerts, and improving efficiency at sea.


Crisp, D. J. (2004). The State-of-the-Art in Ship Detection in Synthetic Aperture Radar Imagery.

De Zan, F. and Monti Guarnieri, A. (2006). TOPSAR: Terrain Observation by Progressive Scans. IEEE Transactions on Geoscience and Remote Sensing.

ESA. (n.d.). Sentinel products specification document. Retrieved from

ESA. (n.d.). Sentinel-1 instrument payload. Retrieved from

ESA. (n.d.). SNAP Download. Retrieved from

Graziano, M. D. et al. (2012). Constellation analysis of an integrated AIS/remote sensing spaceborne system for ship detection. Advances in Space Research.

Kurekin, A. A. et al. (2019). Operational Monitoring of Illegal Fishing in Ghana through Exploitation of Satellite Earth Observation and AIS Data. Remote Sensing.

Lorenzzetti, J. A. et al. (2010). A Performance Comparison Of A CFAR Ship Detection Algorithm Using Envisat, RadarSat, COSMO-SkyMed and Terra SAR-X Images. SeaSAR.

NATO Shipping Centre (2021). AIS (Automatic Identification System) overview. Retrived from,size%20to%20carry%20AIS%20onboard

Paes, R. L. et al. (2010). Ship Detection Using TerraSAR-X Images in the Campos Basin (Brazil). IEEE Geoscience and Remote Sensing Letters.

Sandau, R. et al. (2010). Small satellites for global coverage: Potential and limits. ISPRS Journal of Photogrammetry and Remote Sensing.

Soldi, G. et al. (2020). Space-Based Global Maritime Surveillance. Part I: Satellite Technologies. IEEE Aerospace and Electronic Systems Magazine.

Ulaby, F. and Long, D. G. (2014). Microwave Radar and Radiometric Remote Sensing, University of Michigan Press.

Zuhlke, M. et al. (2015). SNAP (Sentinel Application Platform) and the ESA Sentinel 3 Toolbox. Sentinel-3 for Science Workshop.