Description:Field of Invention:
The present invention pertains to the field of marine pollution monitoring and maritime activities generally. More specifically, it concerns a system and method for detecting oil spills in ocean waters using vessel tracking information from the Automatic Identification System (AIS) together with remote satellite observations. The present invention identifies potential for oil spills by monitoring vessel movement and satellite images of the sea surface. Using a combination of vessel tracking information and remotely-sensed images provides a more timely and efficient manner of detecting marine oil discharges, helping to protect the marine environment and complying with environmental laws.

Objectives of the Invention:

This invention highlights the fundamental importance of seeing oil spills early and being able to take action in order to reduce the adverse environmental and economic impacts. The initial goals of the system as described includes identifying abnormal behaviour of vessels through the use of Automatic Identification System (AIS) data that projects an abnormality in vessel movement behaviour or conduct that may define questionable activity with the higher possibility of leading to if properly managed, a smaller environmental catastrophe. Secondly, which may occur simultaneously, is the satellite imagery activity that clearly recognizes oil spills present and can define with spatial reference the marine area impacted by the oil spill. The system was designed to produce notifications and alerts appropriately and sufficiently in advance of reporting to maritime authorities of distress and crucially emergency response units for the oil spill and enabling them to take adequate action in a short timeframe. Furthermore, by correlating the AIS data to the spill event locations the invention can pinpoint the vessel causing the spill event to potentially support regulatory and penal enforcement actions against offending vessel operators.

Background of the Invention:
Marine oil spills are a significant risk to the environment, impacting marine ecosystems, public health, and economies reliant upon fisheries, tourism, or marine commerce. Oil spill detection relies on visual detection by a human which could delay response due to reliance on manual reports or satellite detection delay. As established in German Patent DE102019201988A1, there is a pressing need for systems that can reliably and automatically detect oil spills with decisive advantageous decisive action whilst addressing timely detection issues which are particularly important in areas more vulnerable to spill risks and areas with a high marine traffic and activity.
Many technological improvements have been made to solve this problem. For example, satellite-based detection processing methods using Synthetic Aperture Radar (SAR), optical and infrared sensors have also been developed including the remote sensing classification of marine oil slicks in 2022. However, the methods are limited by the high cost of deployment, interference from weather patterns, and low revisit time which makes it difficult to respond quickly in dynamic ocean processes. A patent US2020/0184153A1, would improve reporting by offering a method that brings together AIS vessel data with satellite imagery, providing real-time detection of spills and tracking of enforcement.
There has also been effort on the use of artificial intelligence applying deep learning to classify oil spills for example in 2021 CNN based model was developed to discriminate between oil spills and more natural ocean features in satellite images. While this model improved classification results, it struggled to discriminate anthropogenic spills from biological slicks or natural patterns in the ocean. In 2022 deep learning was applied to SAR data which provided improved spill modelling but the deep learning algorithm was dependent on a very large annotated data set and high computational architecture resulting in limiting its practical use within a remote monitoring system.
Efforts have also been made to apply artificial intelligence and deep learning to classify oil spills. In 2021, a CNN-based model was introduced to distinguish oil spills from natural oceanic features in satellite imagery. Although this model improved classification accuracy, it encountered challenges in differentiating anthropogenic spills from biogenic slicks or natural ocean patterns. Similarly, the 2022 use of deep learning on SAR data provided better spill mapping but was heavily reliant on large annotated datasets and high computational infrastructure, limiting its practical deployment in remote monitoring systems.
More recently, hybrid methods have been developed. In 2023, anomaly detection algorithms using AIS data were used to classify suspicious vessel movements associated with illegal discharge, but these approaches were unable to verify spills themselves without looking at satellite observation data. In 2024, a new multi-sensor fusion model enhanced anomaly classification by integrating real-time satellite imagery with AIS data and machine learning algorithms; however, these algorithms were unable to consistently emit alerts on spills, as satellite observation data could often take hours to receive and process. In contrast with the general methods described above, the present invention offers a novel system or framework that synchronizes AI processing for real-time AIS streams with the satellite observation data, enabling timely identifies to be sent out on whether a spill exists, and the regulatory agencies send vessels attribution alerts to follow up.
Summary of the Invention:
This invention combines Automatic Identification System (AIS) vessel detection data with satellite images to create an integrated system for detecting oil spills in marine environments. The system detects vessels exhibiting atypical behaviour signalling illegal discharge, obtains satellite measurements to verify that oil is present on the ocean surface using remote sensing technology, creates a warning to the submitter (relevant maritime authority) to assist in a quick response to investigate oil containment using the AIS report as a starting point, and finally the system can identify the vessels as responsible for the discharge providing evidence to the authorities for enforcement of environmental protection regulations.

Detailed Description of the Invention:
The introduction covers a strong system that starts with anomalous detection from an AIS perspective. The Automatic Identification System tracks the history of vessels, including speed, direction, and unexpected stopping. All of this behaviour information is then processed through machine learning algorithms and analysed for observations that can indicate possible illegal oil discharge. The system was trained to recognize anomalous behaviour such as unexpected slowdowns or stops in open water, which would indicate an illegal discharge. The analysis is done using the vessel's current behaviour compared to historical trajectory data of the same vessel or similar vessels to reduce the chance of false positives, and to increase validity of all flagged behaviours.
Beyond its detection and alerting functionality, the system benefits from a further feedback-based Learning system that will improve model performance over time. Whenever new spill events are verified by authorities or a responsible onsite team, the related images, AIS data, and response results, are added to the growing labelled dataset. The dataset is then utilized to retrain the machine learning and image recognition models on a periodic basis. Thus, the system can adapt to the changing behavioural patterns of vessels and changing environmental conditions, but it also grows better at being able to distinguish between actual spills and falsely detected spills for future events. This continuous learning process helps provide long-term system accuracy and reliability across different marine zones.
After assessing vessel anomalies the next component begins processing and interpreting satellite imagery. This component accesses satellite based Synthetic Aperture Radar (SAR) and optical images covering the square nautical area where the anomaly was identified. SAR images are particularly useful since they can perform readings with cloud cover and in all weather scenarios. Once collected the images are processed through segmentation and thresholding techniques to identify surface anomalies (e.g. oil slicks). The system also uses spectral profiling to identify changes in reflectance associated with hydrocarbons to enhance the effective detection.
Further, the system features a modular architecture to allow seamless integration with external maritime surveillance and government environmental monitoring platforms and fleet management systems. Using RESTful APIs and secure data-sharing protocols, the system can share information regarding real time vessel movements, weather forecasts, and known concentrations of chemical pollutants, and build a complete construction of maritime intelligence. In addition, the system can generate full incident reports with additional metadata such as the time the spill was confirmed, image analysis summary, the identity of the vessel (by its MMSI number), the jurisdictional location (e.g., EEZ boundary), and an estimate of the volume of spill. Incident reports can be archived, exported or presented as evidence during environmental compliance investigations or litigation.
To increase classification accuracy, the system additionally uses a combination of AI-based image analysis and pre-trained models, which are used to differentiate between actual oil spills and natural phenomena that may have a similar appearance (e.g., algal blooms, seaweed mats, or calm patches of water). The models are developed with a portion of the confirmed oil spill imagery that has been curated, allowing the models to recognize natural processes and lookalikes. The models apply advanced feature extraction techniques to identify oil sheen from other anomalous events on the ocean surface. This ensures that the likelihood of oil being present is higher than the likelihood of there being an error due to misidentification of oil slicks. The use of advanced methods on fully labelled data helps ensure that only high confidence detections during classification will proceed to the alert phase.
When the system has confirmed that it has observed a spill, the system initiates the alert and response system. The system automatically issues alerts immediately after detection occurs and communicates information including the GPS coordinates, estimated spill area, time of detection, and likely responsible vessel. The appropriate authorities such as coastal guards, maritime enforcement agencies and environmental disaster response teams are the recipients of these notifications. The system utilizes existing maritime reactive emergency systems for communication channels in advance, allowing quick containment measures to take place.
To maximize classification accuracy, the system provides an AI-based image analysis step leveraging pre-trained models. The models are built utilizing a cleaned set of confirmed oil spill and allow them to differentiate oil spills from natural characteristics or features that resemble oil spills including algal blooms, seaweed mats, or calm patches of water. The platform also provides continuous response via spill tracking and forecasting. It can predict how the spill would likely spread over time by using real-time data of ocean currents, wind speed and direction, then inform spill response managers proper allocation of resources during the spill to mitigate ecological impact. The system will maintain a centralized database event that captures all recorded spill event detections, vessel info, satellite imagery, and any response event timeline. This would benefit agencies to conduct enforcement actions and also help scientists to assess future analysis incidents and also to train the systems.
The system is proven to be useable in high risk operating marine zones like the coastal zone near Mumbai and the Gulf of Mexico, where commercial activity is occurring in a greater density and the likelihood that someone will have unintentional or illegal discharges of oils is significantly increased. The invention combines real time AIS information and satellite based remote sensing and intelligent analysis resources or tools to create a system for marine oil spill detection that is more complete, scalable and automated. The ability of the system to provide advance notice and trigger enforcement is beneficial not only from the environmental protection perspective, but from the safety at sea and law and order compliance and enforcement perspective for international maritime operating activities.
Brief description of Drawing:
Figure 1: Process of AIS-Based Anomaly Detection for Oil Spills.
Figure 2: Integration of Satellite Imagery with AIS Data for Oil Spill Verification and Response. , Claims:The following claims define the scope of the invention:
Claims:
1. A system for detecting oil spills in marine environments, comprising:

a. The AIS-based monitoring module configured to continuously collect and analyze vessel movement data, including speed, direction, and halts.
b. An anomaly detection engine utilizing machine learning algorithms to identify suspicious vessel behavior indicative of potential oil discharge.
c. A satellite data interface adapted to acquire Synthetic Aperture Radar (SAR) and optical imagery corresponding to the detected anomalies.
d. An image processing unit employing artificial intelligence models to analyze satellite images and distinguish oil spill patterns from natural oceanic features.
e. An alert mechanism configured to generate and dispatch real-time notifications to maritime authorities, including spill coordinates, severity, and suspected vessel identity.
f. A blockchain-based logging system designed to assign a unique transaction hash to each verified spill event and store associated metadata for secure, transparent, and auditable tracking.

2. A system as claimed in claims 1, wherein the anomaly detection engine compares current vessel movement patterns to historical AIS trajectory data to minimize false alarms.

3. A system as claimed in claim 1, wherein the satellite data interface is designed to work with multiple satellite sources to improve spatial and temporal coverage.

4. A system as claimed in claim 1, wherein the image processing unit incorporates segmentation, thresholding and spectral profiling modules to isolate and classify surface anomalies.