Answers are grounded in this paper and cite their sources.
In anti-submarine warfare, high resolution, broadband sonars are commonly used. This, however, might challenge the sonar operator in littoral environments as the sonar image often becomes cluttered due to false alarms. The origin of these false alarms can be anything from topographical features in general to man made objects, such as pipelines. The large amount of false alarms in littoral waters makes correct classification of contacts difficult. Normally, a fixed detection threshold on signal-to-noise ratio is employed in order to reduce clutter and make classification feasible. For continuous wave echoes, the same fixed threshold method can be used on Doppler to reduce the false alarm rate. However, this will also lower the probability of detecting a valid target, such as a submarine. A better method for reducing the false alarm rate with a lower reduction of detection probability, is by applying a variable threshold that is dependent on both SNR and Doppler, simultaneously. Here we apply different dependencies between SNR and Doppler in order to find the best target to false detection ratio possible. Due to the classified nature of sonar data with submarines, the dataset used is synthetic with several synthetic submarines present. The various dependencies will be tested on several different synthetic submarines (small, large, fast, slow) to verify that they are not also submarine dependent, as this would reduce the usefulness of the SNR-Doppler dependency significantly. We find that using thresholds that consider both SNR and Doppler simultaneously will in most cases perform much better than a single fixed threshold. Furthermore, the effect on using SNR and Doppler dependent thresholding before clustering has a huge impact on the presented sonar image, improving the amount of false tracks greatly. Finally, we find an unexplained dip in false alarms at a narrow Doppler interval in our synthetic data. This dip will most likely benefit simpler thresholding methods more than the more advanced. Hence, this dip should be understood before recomending using the simpler fixed SNR and Doppler method in false alarm rate reduction.
No takes yet. Share an insight, caveat, or question.
Stender et al. (2019) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: