Automating In-Season Salmon Species Composition Estimation at Mission Using Imaging Sonar

In this project, we will make efforts on improving the accuracy and robustness of automatically measured fish length based on sonar images, since length data can be fitted to a mixture model to determine the species proportion, as currently implemented by Pacific Salmon Commission (PSC) with manually measured fish length data. The mixture model will be added to the existing software package (developed with support by the 2010-2011 Southern Fund). Species composition outputs from this automated software processor can replace PSC’s manual method to significantly reduce the processing time for in-season estimation.

The proposed project will deliver the following 1) automatically measured fish length from DIDSON image data. This measurement function is expected to be available to processing PSC’s ARIS image data providing the ARIS file format is available to ViTech; 2) a mixture model implemented in a stand-alone software program. 3) improved accuracy in species classification between salmon species and resident fish; 4) maximized automation of data processing and improved accuracy in species classification among salmon species.

S18-FRP19 Automating In-Season Salmon Species Composition Estimation at Mission using Imaging Sonar