Open Collections

Edge detection for the segmentation of salmonids in digital images

University of British Columbia

Video footage of salmonids i n sea cages was obtained from a number of areas in Johnstone and Georgia Strait. The footage represented a large number of camera orientations including camera image axis perpendicular and parallel to the water surface. Perpendicular orientations included camera image axis pointing towards the bottom and surface of the water column. Parallel orientations included camera image axis pointing towards the center and side of the sea cage. The use of a white tarp when the cameras were positioned parallel to the water surface provides separation under lower signal to noise ratios. A discussion of noise sources provides a key to understanding the difficulties applying image processing to noisy digital images. In addition, attenuation and scattering when combined with countershading of salmonids illustrates the reflectivity image degradation of salmonids in the water column when viewed from the camera position. Edge detection techniques were applied to noisy digital video images. A comparison of different edge detectors and pre-processing steps was made to determine the "best" separation technique. The Canny operator with a sigma of 2.0 proved to be the "best" edge detector. In order to smooth noisy images, a pre-processing step of Anisotropic Diffusion was applied when the signal to noise ratio was low. This work provides a basis for the automation of counting and sizing assessment of salmonids.

Thesis/Dissertation

application/pdf

2009-01-12

For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use.

http://hdl.handle.net/2429/3586

Biomedical Engineering

University of British Columbia

UBCV

DSpace