Tracking Sponge Size and Behaviour with Fixed Underwater Observatories
More and more fixed underwater observatories (FUO) are being used for high temporal coverage and resolution monitoring of specific areas of interest, such as coral reefs. FUOs equipped with fixed HD cameras and other sensors make it possible to relate visual characteristics of a species to characteristics of its environment.
The aim of this study is to demonstrate how changes in size of sponges can be tracked automatically over time and to investigate to what extent various environmental parameters might influence the behaviour. Both gradual long-term changes and major short-term changes of the sponge size should be taken into consideration. This is the first study that estimates sponge sizes over a long time period.
We present and evaluate an automated process based on a convolutional network (the U-Net) to automatically generate a series of sponge sizes from an image series showing one sponge over a time-interval of 9 month. Further, we analyze the resulting time series together with additional data of the sponge habitat. Our experiments show that our work-flow produces a reliable segmentation of the sponge that can be used for further analysis of the sponge behaviour. Moreover, our results indicate that an increased salinity of the surrounding water is associated to a more frequent pumping activity of the sponge.
KeywordsMarine environmental monitoring Sponge Segmentation CNN
We thank Equinor for the financial support and for providing image and sensor data as well as NVIDIA Corporation for donating the GPU used in this project. We thank Jens Plettemeier for conducting experiments.
- 3.Bernhardt, S.P., Griffing, L.R.: An evaluation of image analysis at benthic sites based on color segmentation. Bull. Mar. Sci. 69(2), 639–653 (2001)Google Scholar
- 4.Ocean Networks Canada: NEPTUNE in the NE Pacific (2014). http://oceannetworks.ca/installations/observatories/neptune-ne-pacific/
- 7.Hoer, D.R.: The role of marine sponges in carbon and nitrogen cycles of coral reef and nearshore environments. Ph.D. thesis, The University of North Carolina at Chapel Hill (2015)Google Scholar
- 9.Kingma, D.P., Ba, J.: Adam: a method for stochastic optimization. arXiv preprint arXiv:1412.6980 (2014)
- 12.Osterloff, J., Nilssen, I., Järnegren, J., Buhl-Mortensen, P., Nattkemper, T.W.: Polyp activity estimation and monitoring for cold water corals with a deep learning approach. In: 2016 ICPR 2nd Workshop on Computer Vision for Analysis of Underwater Imagery (CVAUI), pp. 1–6. IEEE (2016)Google Scholar
- 14.Osterloff, J., Nilssen, I., Nattkemper, T.W.: Computational coral feature monitoring for the fixed underwater observatory LoVe. In: Proceedings of IEEE OCEANS 2016 (2016)Google Scholar
- 15.Ronneberger, O., Fischer, P., Brox, T.: U-Net: convolutional networks for biomedical image segmentation. CoRR abs/1505.04597 (2015)Google Scholar
- 19.Vardaro, et al.: A Southeast Atlantic deep-ocean observatory: first experiences and results. Limnol. Oceanogr.: Methods 11, 304–315 (2013)Google Scholar