261
Manual Counts
In most cases, except when dealing with exceptionally large flocks of birds, a manual count is the most efficient and accurate option, and this will always be the case
if there is a complex background to the image, such as vegetation or rock. For this
reason, manual counts are likely to be the only feasible option for counting seabird
colonies on cliff-faces, for example. A manual count is also the only option if the
birds in the image have a strongly contrasting plumage pattern, such as Avocets
(Recurvirostra avosetta) or Shelducks (Tadorna tadorna) as, in this situation, the
automatic counter will typically count each black section (or white section in an
inverted image) as a separate object/bird.
Manual counts are the default option because they need virtually no preparation.
ImageJ keeps running totals for up to eight species (I have yet to come across a situation where I needed all eight) until the count is complete, at which point there is an
option to save totals to an Excel file. Manual counts in ImageJ have a number of
advantages over carrying out counts in the field and over automatic counts in
ImageJ, including the ability to:
• Zoom into the image and systematically count all of the bird in your own time.
• Pick up the count where you left it if you are disturbed or distracted: this is not
an option if you are disturbed while counting birds in the field.
• Stop to rest your eyes, or for a comfort break, and carry on later when you are
rested and refreshed.
• Decide what should count as one bird, what is part of a bird, what is several
overlapping birds etc.: overlapping birds are counted as ‘one’ in automatic
counts.
• Keep running totals for several species at the same time.
• Commit only one person to the task: it would usually take at least two people to
carry out a seabird population count in the field.
As a general guide, a manual count of c. 1000 birds in ImageJ will take 10–15 min:
Case Study 1 provides a worked example of how to do this.
Manual Count Validation
There is no need for validation of manual counts in ImageJ as you can export images
with markers showing (a) which birds were counted and (b) which birds were allocated to which species (Fig. 10b). If validation is necessary, an independent expert
could check the ‘marked’ image for missed birds or mistaken identification.
Improving the Accuracy of Bird Counts Using Manual and Automated Counts…
Manual Counts
In most cases, except when dealing with exceptionally large flocks of birds, a manual count is the most efficient and accurate option, and this will always be the case
if there is a complex background to the image, such as vegetation or rock. For this
reason, manual counts are likely to be the only feasible option for counting seabird
colonies on cliff-faces, for example. A manual count is also the only option if the
birds in the image have a strongly contrasting plumage pattern, such as Avocets
(Recurvirostra avosetta) or Shelducks (Tadorna tadorna) as, in this situation, the
automatic counter will typically count each black section (or white section in an
inverted image) as a separate object/bird.
Manual counts are the default option because they need virtually no preparation.
ImageJ keeps running totals for up to eight species (I have yet to come across a situation where I needed all eight) until the count is complete, at which point there is an
option to save totals to an Excel file. Manual counts in ImageJ have a number of
advantages over carrying out counts in the field and over automatic counts in
ImageJ, including the ability to:
• Zoom into the image and systematically count all of the bird in your own time.
• Pick up the count where you left it if you are disturbed or distracted: this is not
an option if you are disturbed while counting birds in the field.
• Stop to rest your eyes, or for a comfort break, and carry on later when you are
rested and refreshed.
• Decide what should count as one bird, what is part of a bird, what is several
overlapping birds etc.: overlapping birds are counted as ‘one’ in automatic
counts.
• Keep running totals for several species at the same time.
• Commit only one person to the task: it would usually take at least two people to
carry out a seabird population count in the field.
As a general guide, a manual count of c. 1000 birds in ImageJ will take 10–15 min:
Case Study 1 provides a worked example of how to do this.
Manual Count Validation
There is no need for validation of manual counts in ImageJ as you can export images
with markers showing (a) which birds were counted and (b) which birds were allocated to which species (Fig. 10b). If validation is necessary, an independent expert
could check the ‘marked’ image for missed birds or mistaken identification.
Improving the Accuracy of Bird Counts Using Manual and Automated Counts…