dc.description.abstract | During the last years, Radian AS has been working on new methods for improving the detection
capability of the standard civil marine radar. A coherent-on-receive radar demonstrator
with a PC-based platform has been developed and the methods are found well
suited for detecting small moving targets over land and sea. An operational birdstrike
(aircraft-bird collision) avoidance radar system based on this technology has recently been
deployed at Værnes Airport, Norway. This system uses a slightly modified civil marine
radar, a computer and methods of detection based on a Moving Target Detector (MTD)
processor. The resulting radar video is broadcast to the airport s Air Traffic Control (ATC)
Tower to allow initiation of precautionary measures. Since the current system demands
manual interpretation and constant monitoring of the MTD radar video, there is need for
an Automatic Detection and Tracking (ADT) system and a warning system that draws attention
to specific situations.
In this Master s Thesis, methods for radar detection, tracking and Early Warning (EW)
of avian targets at airports are investigated. The work is based on theoretical analysis, testing with real radar measurements and simulation that incorporates real measurements. The
methods of detection are improved by modification of the MTD processor. A specialized,
batch-processing tracker called a Bird Flight Path Detector (BFPD) Tracker is developed
and implemented to automatically identify and track birds in the airport vicinity. An EW
functionality is also developed and implemented to monitor the resulting tracking data and
give warning of potentially hazardous situations in advance. Furthermore, the performance
of the proposed tracker and the resulting total system is optimized, analyzed and evaluated.
The detection capability of the radar is found sufficient for use in a birdstrike avoidance
application. According to performed theoretical calculations, the existing radar system is
able to detect a single goose at about 4 km with a probability of detection of P_d = 0.7 and
a probability of false alarm of P_fa = 0.001. Testing shows that in practice, multiple flocks
(of varying numbers) of geese are detected consistently enough to allow continuous tracking
by the BFPD Tracker up to about 4 km in range over both land and sea. It is also shown
that the BFPD Tracker is be able to identify and follow all of the important bird presence
while simultaneously exhibiting a probability of true (caused by birds) confirmed track
establishment around 70% and a probability of true batch association around 96-100%.
The latter is hence a good indicator of true bird presence. Simulation experiments show
that the total system is able to detect an avian target roughly the size of a single goose at
ranges of about 3 km with P_d = 0.875 and P_fa = 0.001. Simulation also shows that the
BFPD Tracker is able to track this target continuously up to 4 km over sea with an RMS
error of 2.37 m in range, 0.084 in azimuth and 1.64 m/s in velocity.
The EW functionality is found capable of identifying and giving warning of almost all
manually identified potentially hazardous situations while showing a very low probability
of false warning (<< 1%). Long-term testing and corresponding knowledge of the true
bird activity is needed to accurately estimate the probability of false warning, but this
work indicates that the BFPD Tracker and EW function is suited for tracking and EW
application in an ATC Tower. Near real-time processing is deemed feasible with standard
computing hardware and if the system is developed further it may help mitigate overall
birdstrike risk and contribute to improved safety in aviation. | |