Automatic detection of boats in underwater recordings
Ladda ner
Publicerad
Författare
Typ
Examensarbete för masterexamen
Master's Thesis
Master's Thesis
Modellbyggare
Tidskriftstitel
ISSN
Volymtitel
Utgivare
Sammanfattning
The research addresses a growing environmental challenge in Swedish coastal waters,
where increasing recreational boat traffic poses risks to marine ecosystems through
underwater noise pollution [1]. While commercial vessels can be tracked through
AIS transponders, the vast majority of recreational boats operate without monitoring
systems [2], leaving a significant gap in our understanding of their environmental
impact. To address this challenge, an automated detection system using underwater
acoustic recordings was developed to identify and quantify recreational boat presence
in coastal waters.
Building on established signal processing techniques, a detection system based on
the short-time-average/long-time-average (STA/LTA) algorithm was implemented
[3]. The focus is on optimizing three key components: bandpass filtering parameters,
window functions, and detection thresholds. During fieldwork in August
2024, continuous underwater recordings along the Swedish coast were collected using
SoundTrap and RTsys hydrophones. To validate the system’s performance, Reciever
Operating Characteristic (ROC) curves were used [4] to fine-tune the detection
parameters for optimal sensitivity while minimizing false positives.
Through systematic testing, an STA of 3 minutes and an LTA with the same length
as the recording were found to yield the best performance. The frequency band of
100 Hz to 4000 Hz showed the highest effectiveness in detecting vessel noise.
Threshold optimization revealed that values below 0.5 minimize missed detections, while
values above 0.8 reduce false positives, establishing an optimal operating range of
0.5 to 0.8. These values were obtained from recordings containing numerous boat
passages. They may vary for recordings with fewer passages.
The resulting energy detector offers a practical and adaptive solution for monitoring
recreational boat activity. These findings enhance the ability to quantify
anthropogenic noise in marine environments and provide a foundation for informed
environmental protection efforts.
Beskrivning
Ämne/nyckelord
boat detection, underwater acoustics, energy detection, signal processing, environmental monitoring, recreational vessels, coastal waters, noise pollution