262
Automatic counts
Automatic counts are best suited to situations where, for practical purposes at least,
there are simply too many birds in the image to consider counting manually. These
situations tend to arise in only a few circumsances in the UK, typically at winter
roosts of waders and Starlings. However, the image will need a plain background
and validation is essential.
Automatic Count Validation
Automatic counts come with three options for validation:
• A drawing showing the numbered outlines of objects that contributed to the
count;
• An Excel file documenting the pixel size of every numbered object in the
drawing; and
• The option of carrying out a manual count of all or part of the image.
Validation is an essential source of confidence when applying the automatic
counter and the case studies provide examples of each validation option.
Results
The six case studies focus on situations where it would be difficult to carry out an
accurate count in the field. In the case of automated counts, the accuracy depends
not only on having a suitable image but also on selecting the most appropriate settings for threshold and object size. The case studies below illustrate how to adjust
these settings to best effect for different image types.
A Manual Count of Guillemots at a Breeding Colony at Elegug
Stacks, Pembrokeshire, UK
Counting seabirds at breeding colonies is fraught with difficulties and frustrations.
The number and density of birds being counted is typically high, the birds are constantly coming and going, the birds engage in a lot of territorial disputes, and there
are occasional incursions by predators causing unrest or, even worse, dreads when
many birds will leave the colony etc.
A big advantage of carrying out photographic counts is that you can wait until most
of the birds are present on the colony, take the photo and go home to do the count, safe
in the knowledge that the birds are not going to be disturbed or flushed by predators
part of the way through the count. Text Box 1 details the process used to carry out a
C. Hurford
Automatic counts
Automatic counts are best suited to situations where, for practical purposes at least,
there are simply too many birds in the image to consider counting manually. These
situations tend to arise in only a few circumsances in the UK, typically at winter
roosts of waders and Starlings. However, the image will need a plain background
and validation is essential.
Automatic Count Validation
Automatic counts come with three options for validation:
• A drawing showing the numbered outlines of objects that contributed to the
count;
• An Excel file documenting the pixel size of every numbered object in the
drawing; and
• The option of carrying out a manual count of all or part of the image.
Validation is an essential source of confidence when applying the automatic
counter and the case studies provide examples of each validation option.
Results
The six case studies focus on situations where it would be difficult to carry out an
accurate count in the field. In the case of automated counts, the accuracy depends
not only on having a suitable image but also on selecting the most appropriate settings for threshold and object size. The case studies below illustrate how to adjust
these settings to best effect for different image types.
A Manual Count of Guillemots at a Breeding Colony at Elegug
Stacks, Pembrokeshire, UK
Counting seabirds at breeding colonies is fraught with difficulties and frustrations.
The number and density of birds being counted is typically high, the birds are constantly coming and going, the birds engage in a lot of territorial disputes, and there
are occasional incursions by predators causing unrest or, even worse, dreads when
many birds will leave the colony etc.
A big advantage of carrying out photographic counts is that you can wait until most
of the birds are present on the colony, take the photo and go home to do the count, safe
in the knowledge that the birds are not going to be disturbed or flushed by predators
part of the way through the count. Text Box 1 details the process used to carry out a
C. Hurford