The automatic detection of dark vessels (i.e. vessels that do not report their position to the Automatic Identification System (AIS)) from satellite imagery is particularly challenging in terms of technical capabilities, availability of satellite imagery, satellite revisit time and data availability. In addition, the correct identification of a specific suspicious vessel in a satellite image depends on the ability to identify it using different types of sensors (e.g. radar and optical), as each has different capabilities.
In PERIVALLON, the detection of dark vessels is a task of paramount importance and will be crucial for the Pilot Use Case 3 (PUC3) on “Transnational Illegal Trafficking of Waste Electronic and Electrical Equipment”. The design of this PUC3 is based on a simulated scenario for automatic dark vessel detection using multiple satellite imagery, historical AIS data and background information on some of the main routes for e-waste (electronic waste) trafficking. This scenario has been designed to assist law enforcement agencies in detecting and determining the approximate location of a dark vessel that may be carrying e-waste.
Automatic vessel detection is critical for identifying cargo vessels, particularly when AIS is switched off and they go “dark”, making them difficult to track, most likely with the aim of hiding any potential illegal activity. The use of multi-temporal satellite imagery, based on the Copernicus Sentinel-1 mission Synthetic Aperture Radar (SAR) satellite sensor (Figure 1) from the European Copernicus programme, supports the automatic identification of the vessel of interest (VOI) using computer vision algorithms based on the length and width of the VOI (developed by POLIMI), which also allows the estimation of the gross tonnage (relevant for determining the type of vessel and whether or not it is required to transmit its AIS signal).
Moreover, a historical record of previous AIS reporting periods (provided within the framework of this PUC3 by Kpler) was used to validate the last known position of the vessel before it went “dark”. This information is then used to select the satellite imagery required to track the vessel. From the subsequent satellite images, and in order to reduce the detection threshold, the potential location of the vessel is correlated with AIS data (Figure 2 A and B), which means that vessels that are currently transmitting are automatically removed from the model, reducing the total number of probable vessels of interest identified in a satellite image.
Thus, the combination of automatic detections between different Sentinel 1 satellite images from different dates is used to track the VOI. This approach requires greater technical development and capability, not only to track a dark vessel between different satellite images, but also to provide a higher confidence in the vessels detected and identified by a series of satellite images.
In the future, with the capabilities of the recently launched Sentinel-1C, it is most likely that PUC3 will take advantage of the new capabilities of this satellite as it is equipped with an AIS signal antenna. This new capability will allow non-cooperative vessels to be detected with new tracking and identification capabilities. The aim is to support the decision-making process and reduce the threshold for detecting “dark vessels”.
Written by Juan Francisco Romero Quesada, Jose Santos
SatCen