338 



Fishery Bulletin 89(2). 199) 



• >. 



msam. 



Mm 



Figure 1 



Photograph of a weakfish scale with 29 circuli (29 circuli + 

 edge = count of 30). Arrow depicts counting transect from 

 focus to perimeter along the anterior growing side (transect 

 line = 0.5mm). 



ages of the three counts from individual scales) were 

 then regressed against the visual counts. 



Results and discussion 



Application of a nine-pixel smoothing interval on the 

 transect data combined with a 20-pixel local-minima 

 search width, produced the highest coefficient of deter- 

 mination (r 2 0.99) between the automated and visual 

 counts (Fig. 2). Other combinations of smoothing in- 

 terval and search width resulted in lower coefficients. 

 The present program allows adjustment of smoothing 

 interval and search width to optimize its use with other 

 fish species. Computer counting was approximately 3.3 

 times faster than visual counting. The visual method 

 showed a slightly higher precision compared with com- 

 puter counting, but savings in time more than compen- 

 sates for this small increase in error (Table 1). In addi- 

 tion, the concentration needed and subsequent fatigue 

 in visual counting compared with computer counting 

 are difficult to measure, yet computer counting was 

 considered much easier by all scale readers. 



Microcomputer systems that can digitize increments 

 from scales, otoliths, and other bony structures have 

 been previously reported and demonstrated 

 potential for automated counting (McGowan et 

 al. 1987). The present system makes an ad- 

 vance over other systems by using the local- 

 minimum method of increment identification. 

 Previous methods usually use threshold-light 

 intensity levels to identify increments that 

 may produce discrepancies between computer 



30 40 50 60 



Visual count 



Figure 2 



Linear regression of computer automated count on 

 visual count of scale circuli from weakfish, Cynos- 

 cion regalis (N 45 fish, r 2 0.99, y = 1.03x - 0.3). 



