Automated detection of a crossing contact based on its Doppler shift

Abstract

The trade-off between false alarm and detection probability is a fundamental challenge in the automated detection of contacts in passive sonar systems. A common approach is the application of highgain processing followed by successive classification criteria. Most classification schemes (e.g., matching of signature) are complex and tailored to specific target types. By contrast, the Doppler effect is readily observed in all contacts with discrete tonals and relative velocity to the receiver. This thesis demonstrates that the Doppler effect can be exploited to improve the detection process by filtering out contacts that do not exhibit these characteristics. Cross-correlation (matched filtering) of contact LOFARgrams with templates generated in situ with Doppler compression/dilation is used to achieve this. Velocity information and the range of the Closest Point of Approach (CPA) are estimated in the process. A detection algorithm is developed in MATLAB, and the radiated acoustic signatures of overflying airplanes are recorded at a ground station. In the analysis of six propeller and four jet airplanes, the program successfully identifies the passage of all six propeller airplanes with four incidences of false alarm, due in one case to a jet airplane. Velocity and range estimates are also within expected values.