Project: Zooplankton Viability and Image Analysis of Non-Rigid Motion
Funded by WHOI Ocean Life Institute
How does a learning machine decide if a zooplankton is alive or dead?
Joint Work with Craig Taylor, WHOI Biology Department
Ballast water transfer
is a major pathway for introduction of invasive aquatic species within
the global coastal oceans. The US Coast Guard is encouraging
ship-owners to develop Ballast Water Treatment (BWT) systems and to
demonstrate the efficacy of the various treatment strategies. However, the proposed assessment of zooplankton viability is conducted with a labor-intensive manual microscope and, consequently, does not generate enough statistics to show clearly that the BWT has been effective. We feel that it is essential that the zooplankton viability assay be automated. We propose to develop a video data recording system along with image analysis software that will allow us to quantify viability in 1/10th the amount of time. We will develop a dark field illuminated system which will image the zooplankton. For detection of motion, we will pursue two strategies:
- Modeling the zooplankton as single deformable objects, where changing properties of periphery hugging active contours, or snakes are quantified as a feature vector which is both scale- and rotation-invariant. If the movement of live zooplankton results in feature vectors that are markedly different from those of dead zooplankton, the two feature classes are easily distinguished by a pattern classifier.
- In the situation where the difference between the two classes are more subtle, we will model the zooplankton as articulated objects, consisting of limbs and joint angles. This will provide a more subtle description of the articulated motion and, consequently, the resulting pattern classifier will be more accurate.
- Blake, A. and M. Isard. 1998. Active Contours, Springer-Verlag, Berlin and New York.
- Kakadiaris, I., D. Metaxas, and R. Bajcsy. 1994. "Active Part-Decomposition, Shape and Motion Estimation of Articulated Objects: A Physics Based Approach," Proc. IEEE Conference on Computer Vision and Pattern Recognition, Seattle, pp. 980--984.
- Tiwari, S. and P. Bhattacharya. 1993. "The Recovery of Non-rigid Motion from Stereo Images," Proc. IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), New York, pp 191-194.
- Shawe-Taylor, J. and N. Cristianini. 2004. Kernel Methods in Pattern Analysis, Cambridge University Press, Cambridge and London.
- Zhang, X., Y. Liu, and T.S. Huang, April 2006. "Motion Analysis of Articulated Objects from Monocular Images," IEEE Trans. Pattern Analysis and Machine Intelligence, 28(4) : 624-636.