0 ECS ANI 2003

Detection, classification & tracking of marine mammals using passive, active and ambient noise techniques: 

Detection, classification & tracking of marine mammals using passive, active and ambient noise techniques John R. POTTER Acoustic Research Laboratory Tropical Marine Science Institute National University of Singapore 12 March 2003

In This Presentation……: 

In This Presentation…… The cetacean detection problem Background of statistical Ambient Noise Imaging (ANI) A new concept; coherent multi-static ANI Humpback song units as sources Approach to coherent ANI processing Preliminary feasibility test Summary With the collaboration & support of many colleagues: Michel André, Eric Delory, Adam Pack, Lou Herman, Joe Olsen, Tan Eng Teck, Koay Teong Beng, Matthias Hoffman-Kuhnt, Venugopalan Pallayil, Mandar Chitre, Siri, Mark, Elia, Caroline,Michael & others from the whale house + many more…

Before we begin…: 

Before we begin… “Acoustics remains the least inefficient means of detecting submarines” (or any other large object in the ocean) [Adm. Boisrayon, 2003] “Technological progress is like an axe in the hands of a pathological criminal” [A. Einstein]

Slide4: 

The cetacean detection problem Background of statistical Ambient Noise Imaging (ANI) A new concept; coherent multi-static ANI Humpback song units as sources Approach to coherent ANI processing Preliminary feasibility test Summary

The cetacean detection problem: 

The cetacean detection problem Acoustic impact from powerful sonars & ship-strikes endangers cetaceans Sonar-related incidents in Mediterranean, Bahamas, Canaries, Baja California Fast ferry ship-strike in the Canary Islands & by commercial shipping in the N. Atlantic (Right Whales) Cetaceans sometimes rest at or near the surface where they are hard to see, do not hear well & are are to detect by sonar against surface clutter

The detection problem: 

The detection problem Cetaceans spend only 5-10% of their time visible at the surface, + hard to see at night & in bad wx Active sonar detection adds to the noise pollution unless we use high frequency => short ranges Ship-borne hull sonars suffer from surface clutter => range further limited Ship-borne sonar gives little time to alter course

Cetacean detection problem: 

Cetacean detection problem Vocalising whales can be detected and tracked with arrays Many proof-of-concept examples using SOSUS, Hawaiian islands missile range, Towed Arrays, etc. BUT what about the non-vocalising whales?……. Active sonar would have to be high-frequency => short range + surface clutter constrained Passive sonar cannot ‘see’ silent whales Ambient Noise Imaging sonar…

Slide8: 

The cetacean detection problem Background of statistical Ambient Noise Imaging (ANI) A new concept; coherent multi-static ANI Humpback song units as sources Approach to coherent ANI processing Preliminary feasibility test Summary

Traditional sonar & ANI: 

Traditional sonar & ANI Active sonar Passive Sonar ANI, a third & novel concept in sonar (grew out of the concept of ‘Acoustic Daylight’ by Buckingham and Glegg)

The first ANI camera: 

The first ANI camera Acoustic Daylight Ocean Noise Imaging System (ADONIS) Research acoustic camera built at SIO in 1994 24 fps, 25-85 kHz

Statistical ANI: 

Statistical ANI ‘Targets’ were neoprene foam panels, 1 x 1 m Range was ~50 m No active insonification, silent targets

Statistical ANI images: 

Statistical ANI images Acoustic Daylight (AD) - 1st-order moment of acoustic intensity Sigma - 2nd-order temporal moment Xcorr - 2nd-order spatial moment

Some typical sources of Ambient Noise: 

Some typical sources of Ambient Noise Weather & geophysics Breaking waves & non-linear wave interaction in the open ocean + surf (near shore) Storms (lightning) Microseismic events Anthropogenic Shipping Sonar Industrial Biological Noises Snapping Shrimp Fish Marine mammals

ANI imaging systems: 

ANI imaging systems Acoustic Daylight Ocean Noise Imaging System (ADONIS) ‘Acoustic Daylight’ based camera, uses the intensity of incoherent received energy to form an image Remotely Operated Mobile Ambient Noise Imaging System (ROMANIS) Uses several statistical imaging techniques based on higher statistical moments and can be used for coherent ANI DSTO, Australia A highly-sparse array designed to use a coherent ANI technique

Remotely Operated Mobile Ambient Noise Imaging System (ROMANIS): 

Remotely Operated Mobile Ambient Noise Imaging System (ROMANIS) 500+ sensors 1.4 m aperture 25-85 kHz bw Fully digital 54 Pentium DAQ computers in FC-AL network 1.6 Gbits/s data flow

Slide16: 

The cetacean detection problem Background of statistical Ambient Noise Imaging (ANI) A new concept; coherent multi-static ANI Humpback song units as sources Approach to coherent ANI processing Preliminary feasibility test Summary

Coherent multi-static ANI concept: 

Coherent multi-static ANI concept If individual transient source events can be localised, we can use them as ‘opportunistic’ sources ‘Opportunistic source’ idea leads to multi-static sonar processing - known from ASW work to exhibit advantages over monostatic sonar Coherent ANI is potentially even more advantageous than ‘conventional’ multi-static because of the distributed nature of the sources A parallel exists in passive radar imaging Silent Sentry, passive surveillance radar - Lockheed Martin Mission Systems (1999) Y Wu & D. C Munson, University of Illinois-multistatic synthetic aperture radar imaging This approach forms part of the current work by Michel André & Eric Delory in designing the WACS system.

Deep-water sources: 

Deep-water sources Shipping Quasi-cw, not sufficiently transient Sonars Oil & Gas exploration Cetaceans Sperm clicks, Blue and Fin grunts, Humpback song Problem statement: Ship-cetacean interaction

Slide19: 

The cetacean detection problem Background of statistical Ambient Noise Imaging (ANI) A new concept; coherent multi-static ANI Humpback song units as sources Approach to coherent ANI processing Preliminary feasibility test Summary

Males “sing” in ‘wintering grounds’Singers may be stationary or moving.“Song” typically lasts 10-15 minutesSource level may be 180-200 dB re 1 mPa: 

Males “sing” in ‘wintering grounds’ Singers may be stationary or moving. “Song” typically lasts 10-15 minutes Source level may be 180-200 dB re 1 mPa Characteristics of Humpback ‘Song’

Song Structure: 

Song Structure Unit (~2 sec) Phrase (~4-8 units with 1-2 s gaps) Theme (~2-6 repeated phrases) Song (~4-8 themes) Units

Some unit types: 

Cow Creaky Door Hiccup Rachet Strain Squeal Some unit types Themes

Slide23: 

The cetacean detection problem Background of statistical Ambient Noise Imaging (ANI) A new concept; coherent multi-static ANI Humpback song units as sources Approach to coherent ANI processing Preliminary feasibility test Summary

The Concept Scenario: 

The Concept Scenario Singer units as multi-static ‘opportunistic’ sonar sources Coherent processing of individual units Complicated by the lack of information about where and when the source occurred Target Receiver Sources

Steps to processing coherent ANI: 

Steps to processing coherent ANI Identify an incoming unit/click at a sparse distributed LBL receive array & estimate it’s source time-series. Beamform/matched field across all LBL elements to estimate the source location & emission time. Predict received signal at LBL elements & remove. Process subsequent data at LBL elements for scattered arrivals using estimated time series as a matched filter as for a multi-static active system; LBL array => volumetric localisation is possible. Guard against incorrectly processing surface & bottom reflections &/or new incoming transients as if they were scattered ‘echoes’ of a previous signal by full/matched field processing & time/energy constraints on possible scattered ‘copies’ of original.

Source estimation: 

Source estimation Requirements The signal is a broadband, temporally-compact signal (=> compact autocorrelation). The range is short enough so that signals are not overly distorted by absorption &/or dispersion that cannot be estimated & corrected (correlation provides a good signal detector). Propagation processing can be as simple as geometric ray-tracing or as complex as matched field as deemed appropriate - decoupled problem Distributed ‘random’ sparse arrays can be used as LBL arrays to perform volumetric beamforming

Resolving received signals: 

Resolving received signals After a direct path transient signal is received, a new detected signal could be: A transient reflected off the bottom/surface/multiple reflection An ‘echo’ reflected off a mid-water target A new transient A decision logic has to be developed to separate these possible classes of sources

Resolving received signals (contd.): 

Resolving received signals (contd.) Physical & biological constraints: ‘Echoes’ must be lower amplitude than the original signal (& decreasing with time) - we need to estimate the target strength & emission beampattern of the sources Surface reflections result in a phase shift and must come from upper half plane in same vertical plane as the source Bottom reflections must arrive from lower half-plane in source vertical plane & conform to known critical angle constraints Surface & bottom reflected paths can be used with ray-tracing to give range in a known bathymetric environment

Back-of-the-envelope calculation: 

Back-of-the-envelope calculation To be able to detect an ANI ‘echo’, DT ≤ SL-(TLt-TS+TLr+N) + G DT = Detection threshold SL ~ Source level ~ 200 dB re 1 mPa @ 1m TLt ~ 66 dB for spherical spreading out to 2 km TS ~ -6 dB for another whale TLr ~ 66 dB for spherical spreading out to 2 km N ~ 40 dB re 1 mPa/√Hz + 33 dB for 2 kHz bw G ~ 10 dB (depends on distortion, DTDf, etc.) DT ≤ 0 dB => a tough detection problem…

Slide30: 

The cetacean detection problem Background of statistical Ambient Noise Imaging (ANI) A new concept; coherent multi-static ANI Humpback song units as sources Approach to coherent ANI processing Preliminary feasibility test Summary

Preliminary feasibility test: 

Preliminary feasibility test Conducted in Hawaii February 2003 in collaboration with TDI & KBMML, Univ. Hawaii (Herman & Pack) Deploy 3 PANDA’s along Maui coast of Auau channel Confined area with Humpback singers Sparse LBL array Calibrate with known singer locations & signals Intent is to detect & track active singers Remove active singers from data and process for ANI

Slide32: 

‘Aquahead’ Equipment & Technologies

Slide33: 

Aquahead gives singer direction to ~ 10 deg. Crude range estimate Locates from up to 8 km away Tracks in as little as 10 min.

Slide34: 

Surface snorkeler guides divers & takes sizes using videogrammetry Rebreather (few bubbles) divers use digital video + rangefinders to record song & obtain ranges Orientation to singer obtained from video

Enhanced PANDA: 

Enhanced PANDA Pop-up Ambient Noise Data Acquisition system Self-contained, compact Acoustic & timed release 20 kSa/s 2 days’ continuous endurance Acoustic modem allows status query & reprogramming without releasing

PANDA distributed random array - transient tracking, localisation & classification: 

PANDA distributed random array - transient tracking, localisation & classification PANDAs Singers

PANDA array processing: 

PANDA array processing LBL array Random positioning Calibrated by known singer unit time-series & GPS location Simultaneous shallow-water, multiple transient Detection Location (random array processing) Tracking Classification (wavelets & Neural nets) Iterative singer ‘removal’ & ANI multi-static coherent imaging

Slide38: 

The cetacean detection problem Background of statistical Ambient Noise Imaging (ANI) A new concept; coherent multi-static ANI Humpback song units as sources Approach to coherent ANI processing Preliminary feasibility test Summary

Summary: 

Summary ANI has been shown to be effective to image silent targets without introducing additional sound into the environment => non-invasive & the only sonar able to silently image silent targets Ranges from statistical high-frequency results (snapping shrimp) are currently very short but ANI using multi-static coherent processing at lower frequencies has potential to image cetaceans over several km. New techniques & enhancements are developing rapidly on several fronts and in several places (e.g. WACS)

Slide40: 

Thank You… For your attention, your humanity & your passion to love & protect our environment