265
Text Box 2: The Process and Settings Used for Carrying Out the
Automatic Count of a Starling Murmuration (Fig. 3) in ImageJ
Open ImageJ, select ‘File’ ➾ ‘Open’ – and choose the image that you want to
analyse
When the image is open select ‘Image’ ➾‘Type’ ➾‘8-bit’
Then, again under the ‘Image’ menu, select ‘Adjust’ ➾‘Threshold’ – this
will open a box containing the Threshold settings
Then select the appropriate Threshold settings which, in the case of the
image of the Starling murmuration, were as follows:
Top slider
– 0
Bottom slider
– 119
Default
B&W
Don’t tick
‘Dark background’
Click on the apply button
Then go back to the main ImageJ toolbar and select ‘Analyze’ ➾‘Analyze
particles’ – this opens another box and the following settings were selected:
Size (pixel)
10-infinity
Circularity
0.00–1.00
Show
Outlines
Tick
Display results
Click on OK
This will generate an Excel spreadsheet with the count details and an
image labeled ‘Drawing of + file name’ this shows the Starling outlines (numbered) that contributed to the count total.
is a drawing showing the numbered outlines of every object counted and a table
showing the area (in pixels) for each object, you can cross-reference between the
two to confirm this and to check whether the objects with a larger pixel area comprised more than one bird. Cross-referencing between the drawing and Excel table
for the Starling image revealed that objects of 100–199 pixels comprised at least
two birds. Therefore, as many objects in the spreadsheet had an area of >100 pixels,
it was clear that the automatic count of 9380 birds must have been an underestimate,
and quite possibly a substantial one.
To improve the accuracy of the Starling estimate, I scrolled through the spreadsheet and for every object with an area of 100–199 pixels I added one bird to the
total, and for every object with an area of 200–299 pixels I added two birds etc., etc.
Improving the Accuracy of Bird Counts Using Manual and Automated Counts…
Text Box 2: The Process and Settings Used for Carrying Out the
Automatic Count of a Starling Murmuration (Fig. 3) in ImageJ
Open ImageJ, select ‘File’ ➾ ‘Open’ – and choose the image that you want to
analyse
When the image is open select ‘Image’ ➾‘Type’ ➾‘8-bit’
Then, again under the ‘Image’ menu, select ‘Adjust’ ➾‘Threshold’ – this
will open a box containing the Threshold settings
Then select the appropriate Threshold settings which, in the case of the
image of the Starling murmuration, were as follows:
Top slider
– 0
Bottom slider
– 119
Default
B&W
Don’t tick
‘Dark background’
Click on the apply button
Then go back to the main ImageJ toolbar and select ‘Analyze’ ➾‘Analyze
particles’ – this opens another box and the following settings were selected:
Size (pixel)
10-infinity
Circularity
0.00–1.00
Show
Outlines
Tick
Display results
Click on OK
This will generate an Excel spreadsheet with the count details and an
image labeled ‘Drawing of + file name’ this shows the Starling outlines (numbered) that contributed to the count total.
is a drawing showing the numbered outlines of every object counted and a table
showing the area (in pixels) for each object, you can cross-reference between the
two to confirm this and to check whether the objects with a larger pixel area comprised more than one bird. Cross-referencing between the drawing and Excel table
for the Starling image revealed that objects of 100–199 pixels comprised at least
two birds. Therefore, as many objects in the spreadsheet had an area of >100 pixels,
it was clear that the automatic count of 9380 birds must have been an underestimate,
and quite possibly a substantial one.
To improve the accuracy of the Starling estimate, I scrolled through the spreadsheet and for every object with an area of 100–199 pixels I added one bird to the
total, and for every object with an area of 200–299 pixels I added two birds etc., etc.
Improving the Accuracy of Bird Counts Using Manual and Automated Counts…