Page 25 - 2024 - Q2 - Minerva in Focus

MINERVA IN FOCUS
MINER V A IN FOCUS Digital innovation in shipping

• Integrated radar networks can cover larger areas, including • Data Integration: Combining data from multiple sources to
international waters. create a unified dataset.
• Sonar systems can detect the presence and movement of • Normalisation: Standardising data formats and units to ensure
submarines and other underwater vessels, which may be consistency across the dataset.
used for smuggling or illegal fishing.
3. Feature Extraction
2. Environmental DNA
• Sampling water for eDNA can help identify the presence of
certain species in specific areas, indicating illegal fishing
activities.

3. Cameras
• Cameras installed on vessels can capture images and vid-
eos of fishing activities, which can be analysed to ensure
compliance with regulations.
4. Port and Docking Data
• Records of port calls and docking activities can be analysed
to detect unusual patterns or identify vessels that avoid of- • Behavioural Features: Features that describe the vessel’s
ficial ports. behaviour, such as speed, direction, changes in course, time
• Cargo manifests and shipping records provide details on spent in certain areas, and proximity to vessels.
the goods being transported, which can be checked against • Geospatial Features: Identifying the vessel’s location relative
other data for inconsistencies. to geographic features.
• Temporal Features: Analysing time-based patterns, such as
So, how does it work? the time of day, season, and duration of activities.
Algorithms that identify illegal actions by vessels operate through
a series of sophisticated processes involving data collection, 4. Pattern Recognition
integration, analysis, and decision-making. • Normal Behaviour Modelling: Using historical data to establish
patterns of normal behaviour or different types of vessels.
1. Data Collection • Anomaly Detection: Employing statistical and machine learn-
ing techniques to detect deviations from normal behaviour.
5. Advanced Analytics and AI
• Deep Learning: Utilising deep learning models to analyse
complex patterns in large datasets, such as identifying specific
fishing techniques from satellite imagery.
• Natural Language Processing: Analysing text-based data
from logs, reports, and social media for additional context
and intelligence.
• Real-Time Data: Continuous collection of real-time data • Fusion Algorithms: Combining outputs from multiple mod-
from sources such as AIS, VMS, satellite imagery (optical els and data sources to improve accuracy and reduce false
and SAR), and coastal radar. positives.
• Historical Data: Aggregation of historical data on vessel
movements, fishing patterns, and enforcement actions. 6. Decision Support and Action
• Auxiliary Data: Inclusion of additional data such as weather • Risk Assessment: Evaluating the risk level associated with
conditions, oceanographic data, port records, and trade data. detected anomalies to prioritise enforcement actions.
• Reporting: Generating detailed reports on suspected illegal
activities, including supporting evidence and recommended
actions.
• Collaboration: Sharing information and alerts with relevant
authorities, such as coast guards, marine patrols, and inter-
national organisations.

2. Data Preprocessing
• Data Cleaning: Removing noise and correcting errors in the
raw data.

23