WP4. Smart Surveillance Algorithms


This work package is responsible for processing and analysis of the sensor data, including radar, video stream and acoustic information. From the video side, it aims to investigate methods to recognize and track ships by appearance without using a model of the background as these conventional techniques are too sensitive for light artefacts. We will study detection by feature-based machine learning and detection by using multi-position (e.g. stereo based) sensing in combination with epipolar geometry. Moreover, we will explore advanced cognitive functions to extract suspicious behaviour in order to provide alert for security issues. Suspicious indicators are not corresponding AIS information, route deviations, predicted collisions, small vessels departing from ships, etc.

For vessel’s safe navigation, automatic route planning algorithm will be analyzed and developed according to work package objectives.

On the acoustic side, above all, suitable sensor packagings will need to be developed and tested that should protect the delicate sensors from the salty water. It should allow the deployment of acoustic vector sensors in a marine environment, the sensor nodes being placed: On the ground

   · In the air, on board of a UAV or LTA airborne surveillance platform

   · At sea, at the sea bottom or on board of a floating sensor buoy


Furthermore, novel algorithms will be developed to dected, locate and track small vessels. The information obtained can be fused with data obtained from other sensors.

 This work package also includes tasks to monitor the following environmental parameters such as pH, conductivity, oxygen, temperature, turbidity. This information is useful to detect environmental disasters or illegal waste disposal. 



Task 4.1: Small Vessel Monitoring


Task leader : GMT

This task will generate an important innovation and comprises research in several sensor modalities since traditional radar system are not capable of detecting small vessel. EO/IR images allow detection of small structure that deviate from the water while directional acoustic sensors are capable of recognizing the noise from engines and propellers. Furthermore, a study on technology to manage small vessel using UHF (Ultra High Frequency) whose bandwidth is 400MHz or 900MHz. In addition, this task includes the development for the module of interface for legacy sensors, such as current, wind, gyrocompass, fire detection, and power detection, and communication, such as UHF, 3G/LTE, AIS, and Wi-Fi, according to Plug & Play architecture profiles defined in Task 3.3.



Task 4.2: High-reliable tracking


Task leader : TKH-Siqura

Small vessels and crowded vessel traffic in harbour ports cause unreliable tracking information from radar posts, especially in the vicinity of land infrastructure such as buildings and bridges. Using different modalities and sensors from different positions and angles we are capable to increase the reliability of the vessel tracks for traffic management and to reduce the cost of surveillance for a certain perimeter or area. 



Task 4.3: Vessel classification, identification and verification


Task leader : ViNotion

Typically identification of vessels is performed via the AIS. However, non-cooperative vessels may transmit false information or not provide information at all. In this task we will study advanced fingerprinting technology using features from EO/IR sensors, directional acoustic sensors, radar and AIS. This will enables re-identification and reliable tracking over large distances active communication with vessels.



Task 4.4: Behaviour analysis


Task leader : TU/e

Apart from reliable monitoring of vessel traffic we aim to automatically recognize suspicious behaviour and abnormalities such as false AIS information, route deviations, predicted collisions, small vessels departing from ships, pollution, etc. This requires, the multi-modal sensor data and ontology based reasoning to infer high semantic-level information for decision making.

This task therefore contains Complex Event Processing (CEP) technologies. The results will be integrated into a Geographic Information System (GIS) to create “geospatial awareness” and to allow rules and conditions for the CEP engine. The rule language allows to add new operators or any other language abstraction in order to allow the developers to express spatial conditions or relationships.