Multi Sensor Data Fusion

Transcription

1 Multi Sensor Data Fusion for improved maritime traffic monitoring in the Canadian Arctic Giulia Battistello*, Martin Ulmke*, Javier Gonzalez*, Camilla Mohrdieck** (*) Fraunhofer FKIE Sensor Data and Information Fusion - Wachtberg, Germany (**) Airbus Defence a - Data Fusion Concepts, Integration & Tests Ulm, Germany 9 th Symposium of the International Society of Digital Earth (ISDE) Halifax Nova Scotia, Canada, 5-9 October 2015

2 PASSAGES PROJECT Protection and Advanced Surveillance System for the Arctic: Green, Efficient, Secure Canadian-German Partnership: July 2013 June 2016 To specify the requirements and the modular architecture of an innovative maritime system to support operations in Arctic waters with a focus on the Northwest Passage

3 Case Studies for Canadian Arctic (I) Problem: the compilation of maritime traffic picture i.e. the basis for safety and security applications is limited by the scarceness/absence of local sensors. Sensor data might be missing due to Sensor limited performance or unavailability Data transmission interruption Data blanking or spoofing Objective: Development of a monitoring service that guarantees an higher update rate of vessel tracks with a minimal deployment of new sensors

4 Case Studies for Canadian Arctic (II) Scenarios under consideration: Choke Point Monitoring (Bellot Strait) Open Waters Monitoring (Resolute) Anomaly Detection and Assessment (Bellot Strait) Search and Rescue (Devon Island) Covert Monitoring (Frobisher Bay & Hudson and Davis Straits)

5 Case Study: Covert Monitoring Scenario Frobisher Bay & Hudson/Davis Straits Frobisher Bay Only waterway to Nunavut s capital Iqaluit 120km length, [20-40]km width with several islands Detailed charts not available Presence of ice Frobisher Bay Davis Strait Hudson and Davis Straits fishing grounds Illegal fishing activities border crossings

6 Input Data Source SAT-AIS Satellite Automatic Identification System Latitude [deg] Time Window: Longitude [deg] Real Data from Exact Earth Vessel ID, Position, Velocity, SOG, COG, Heading, Timestamp, Type and more Cooperative Vessels Vessels > 300t (500t) Measurements come as bursts with inter-bursts of 90 mins Discontinuous tracks Fishing Passenger Cargo Ice Breaker Oil/Chemical Tanker Warship

7 Input Data Source SAT-SAR Images Satellite Synthetic Aperture Radar Real Data from RadarSAT-2 and TerraSAR-X Position, Heading, Length Low satellite revisit frequency for continuous surveillance False detections, in particular by ice :44:02 UTC :13:02 UTC Validated with SAT-AIS Not validated

8 Proposal: Passive Radar as local sensor Sensor principle Multistatic radar using non-cooperative illuminators, e.g. GSM, UMTS, LTE, DAB, FM, HF, VHF Output Object detection Tracks GAMMA-2 system (FKIE) Platforms Antenna arrays installed on hill, tower, lifting platform Potentially aircraft, UAS, airship Weaknesses Availability of broadcasting stations Airbus D&S multiband passive radar 8

9 Input Data Source Passive Radar Simulated Data from network of passive sensors that exploit signals already present in the surveillance area as illuminators of opportunity GSM base stations and/or VHF radio stations Position, Velocity of moving vessels TX1 Advantages Reduced electro magnetic pollution Reduced installation and maintenance costs Not subject to authorization by safety authorities Covert tracking >> detection of non-collaborative vessels (not equipped with AIS or not using it) RX2 TX2 RX1

10 Passive Radar Data Simulation Set Up Analysis of maritime traffic through historical data Identification of available transmitters (Txs) Ground Truth Measurement Errors Optimization of the bistatic passive radar geometry (Tx-Rx) >> maximization of the target detection capability (PD) Sensors Definition AoI Sensors description AoI Sensors description Passive Radar Measurements Simulator Measurements Simulation of passive radar measurements and tracks Generation of Detection Probability Maps PD Maps Target Tracker Tracks

11 Multistatic Geometry Sub-Area Definition (I) 64 Area # Area #2 Latitude [deg] Area # Longitude [deg] Passenger Cargo Tanker TX 6 (4 real 2 new) FoV = 120 RX 1 GSM Field of View = 210 BW = 83.5KHz Angular Res = 3.2

12 Sub-Area Definition (II) 64 Area # Area #2 Latitude [deg] Area # Longitude [deg] Passenger Cargo Tanker TX 6 (4 real 2 new) FoV = RX 1 GSM - 1 VHF Field of View = BW = 83.5KHz 15kHz Angular Res =

13 Sub-Area Definition (III) 64 Area # Area #2 Latitude [deg] Area # Longitude [deg] Passenger Cargo Tanker TX 1 Omnidirectional RX 1 VHF Omnidirectional BW = 15kHz Angular Res = 10

14 Target Tracking Results (I) x Detection Probability Map Y [m] Lat [deg] Ground Truth PR Tracks X [m] x 10 5 Simulation Parameters Target RCS = 100 m2 Height = 20 m MDV = 4 m/s PFA = 10-3 s Lon [deg] Results 5 Targets in the AoI in the time window 42 track segments from Passive Radar

15 Target Tracking Results (II) 6.94 x Detection Probability Map Ground Truth PR Tracks Y [m] Lat [deg] X [m] x 10 5 Simulation Parameters Target RCS = 100 m2 Height = 20 m MDV = 4 m/s PFA = 10-3 s Lon [deg] Results 15 Targets in the AoI in the time window 33 track segments from Passive Radar

16 Target Tracking Results (III) Y [m] x Detection Probability Map X [m] x 10 5 Simulation Parameters Target RCS = 100 m2 Height = 20 m MDV = 4 m/s PFA = 10-3 s -1 Lat [deg] Ground Truth PR Tracks Lon [deg] Results 15 Targets in the AoI in the time window 35 track segments from Passive Radar

17 Data Fusion Engine: Building Blocks Two basic integration levels Sensor & source level integration of measurements from a class of sensors with similar characteristics (e.g. Radar & IFF) and integration of track or navigation data from an external source (e.g. tell unit on a datalink, AIS source,..) Result: sensor level tracks Track to track fusion level Integration of data from sensor level tracks into system level tracks (MSI Tracks = Multi Source Integrated Track) Track to track correlation&association Kinematics and attribute integration Identification and classification

18 Radar - AIS Integration (Example from Baltic Sea) Low AIS reporting rate Both AIS and Radar have good kinematics. MSI creates fused kinematics MSI track AIS track frequent Radar update rate Radar track external AIS track symbol MSI System Track symbol

19 Radar - AIS Integration Bad Radar track kinematics. Good AIS kinematics. MSI creates kinematics based on preferred AIS but interpolates with Radar update rate. MSI track AIS track directly underneath MSI track Radar track MSI System Track symbol

20 Radar - AIS Integration Radar only MSI track AIS reporting starts MSI prefers AIS kinematics Only one AIS update merged to the MSI track. But the AIS information is maintained on the MSI track. See e.g. MMSI number is retained MSI uses Radar kinematics

21 Use Case Data Fusion Results (I) Fusion of SAT-AIS tracks and Passive Radar tracks leads to higher update rate of vessel tracks >> vessel monitored for longer time window (red plots correspond to passive radar only observations) Data availability / P D Data availability / P D Track segmentation?

22 Data Fusion Results (II) Data fusion allows improving track continuity Lat [deg] Ground Truth PR Tracks Fused Tracks SAT-AIS Lon [deg] Target in AoI Visibility PR Tracking Output Data Fusion Output Segments 37 Associated to 5 tracks + 5 Not associated Segments 35 Associated to 10 tracks Lat [deg] Lon [deg]

23 Conclusions Case Study: Covert Monitoring Scenario >> enhanced traffic monitoring by a minimal deployment of new sensors Satellite AIS reports augmented by GSM/VHF based Passive Radar as innovative sensor Non cooperative vessels Fusion of heterogeneous sensor data leads to Current work Higher update rate of vessel tracks Improved track continuity Other sensors: Sat-SAR-detections, coastal radar, Enhanced tracking and route prediction using GIS data (coastlines, bathymetry)

24 Thank you for your attention! Dr. Martin Ulmke Head research group Wide Area Surveillance Department Sensor Data and Information Fusion (SDF) Fraunhofer FKIE Webpage: passages.ie.dal.ca