1. Post-
processing
Pouliquen echo
correction for
• Daniel Rodríguez Pérez, UNED
• Noela Sánchez Carnero, UDC
bottom
• Juan Freire Botana, UDC classification

2. The problem…
• Sea bottom classification
• Muddy
• Sandy
• Rocky
(Folk classification of sediments)
• Single beam echosounder
• Cheap and simple
• Transect based
• Coastal areas
• Fast bottom type changes
• Large depth differences

3. Our goal…
• Use single-beam echosounder
• Correct echoes in coastal areas
• Get a sound classification
• In a coastal area

4. Study area
• Ría de Cedeira
• Acoustic survey
CORERS
• Diving transects
• Fishermen map
• Interview based
• Consensus map

5. The correction and the idea…
• Power correction • Pouliquen ECAS 2004
• +30 log ( R/Rref ) • Change pulse duration online
• Time correction • Our approach
• Time stretching • Change pulse duration offline
• How?
• Convolution

6. The correction and the idea…

7. The correction and the idea…
• Time correction • Pouliquen ECAS 2004
• Time stretching • Change pulse duration online
• Power correction • Our approach
• +30 log ( R/Rref ) • Change pulse duration offline
• How?
• Convolution
Time adjustment Power adjustment

8. The correction and the idea…
• Time correction • Pouliquen ECAS 2004
• Time stretching • Change pulse duration online
• Power correction • Our approach
• +30 log ( R/Rref ) • Change pulse duration offline
• How?
• Convolution

9. The correction and the idea…
• Time correction • Pouliquen ECAS 2004
• Time stretching • Change pulse duration online
• Power correction • Our approach
• +30 log ( R/Rref ) • Change pulse duration offline
• How?
• Convolution
CONVOLUTION

10. First results
• Pings over sand in the study area
• Different depths
• Pings over rock in the study area
• Different depths
• Larger variations

11. Classification
• Ping average 38 kHz
• 20 ping averages
• 50 m segments
• Nearby ping grouping
• Spatial coherence
• Transect classification
• Global ping grouping
• Global classification
200 kHz
• Class representatives
• Ping classification

12. Classification
• Ping average
• 20 ping averages
• 50 m segments
• Nearby ping grouping
• Spatial coherence
• Transect classification
• Global ping grouping
• Global classification
• Class representatives
• Ping classification

13. Classification
• Ping average
• 20 ping averages
• 50 m segments
• Nearby ping grouping
• Spatial coherence
• Transect classification
• Global ping grouping
• Global classification
• Class representatives
• Ping classification

14. Validation…
• Fishermen map validation
• qualitative
• Diving transect validation

15. Validation…
• Fishermen map validation
• qualitative
• Diving transect validation
FS&M CS R
CL1-3 259 8 10
CL2 5 2 40
CL4-5 6 1 157
κ=0.73

16. Discussion
• Ping averaging
• (forced) Spatial coherence
• Physically (intensity) motivated
• What about rocky bottoms?
• Classification based on templates
• (implied) Spatial coherence
• Transducer (and frequency) dependent
• Accuracy and kappa? Not too good
• Ground-truthing samples
• Only one frequency (and one echo per ping) used

17. Conclusions
• PLP correction
• Effectively corrects the shape of the echoes to a shallow water
reference
• … at the cost of losing some details
• Aggregative method classification
• Ping averaging taken to its most
• Echo template based classification
• … relative success
• Further research needed for rocky bottom correction

18. Acknowledgements
• The authors wish to acknowledge Prof. M. Mozynsky for
enlightening discusion about the ideas in this paper
• ... and the organization of the ACUSTICA 2012 for the
opportunity to present this work here.

19. Acknowledgements
• The authors wish to acknowledge Prof. M. Mozynsky for
enlightening discusion about the ideas in this paper
• ... and the organization of the ACUSTICA 2012 for the
opportunity to present this work here.