USING IMAGE ANALYSIS OF ACOUSTIC DATA



 
 
Patricia LeFeuvre*, George Rose+, Ray Gosine*0, Rafaat Khan*0, Chris Pike * 
* C-CORE - Center for Cold Ocean Resources Engineering
lefeuvre@engr.mun.ca

+ Senior NSERC Chair in Fisheries Conservation
Fisheries and Marine Institute of Memorial University of Newfoundland

0 Faculty of Engineering and Applied Science
Memorial University of Newfoundland


 


Quantifying the distribution and abundance of fish in their natural habitat with little disturbance is the greatest strength of acoustic technology. 
A major problem with acoustic methods is identifying the echoes. Traditionally, surveys relied on fishing to identify the fish observed. However, fishing results are affected by the catchability of each specific species and the resolution of fishing data is never comparable to that of high resolution sonar. For years fishermen and experienced researchers have distinguished between different species based on echogram images. Continuing advances in fisheries science, acoustic technology, digital signal/image processing have made the possibility of semi-automated fish classification a more realistic goal. 

We are currently developing a toolkit for species identification and acoustic data analysis. The toolkit is a 32-bit Windows application that will run on Win NT or Win95. At present the data is collected using two Biosonics DT4000 systems: a 120 kHz and a 38 kHz; other sensors will also be brought into play. The software will be used to generate data sets of features for cod, capelin and redfish. The echogram below is an example of output from this software. 

DT 4000 Echogram showing Cod in 3PS
The above image is taken from the new Fish Species Indentification Software (FSIS) copyright 1996.