455
In parallel, many whale populations have been recovering, including southern
right whales (IWC 2013) and so the number of sightings and consequently individuals photographed in each aerial survey. This demanded for the development of some
automated system which could assist the comparisons. In 1998, Hiby and Lovel
(2001) developed a computer program called “Right Whale Photo-identification
Software”, to aid on the analysis of aerial photographs taken from right whales.
The “Right Whale Photo-identification Software” makes use of the variation in
shape, number and location of callosities on the rostrum and lips, transforming the
three-dimensional shape of the head of right whales in a two-dimensional picture,
where image pixels define the presence of callosities, generating a library of patterns for comparison with new images analyzed. The objective of the software is to
suggest the order in which the existing catalogue be searched for matches to a new
photograph, and to eliminate from the search those photographs in the library that
cannot possibly be potential matches of the new one (Hiby and Lovell 2001).
The method consists in extract from each new photograph a simplified map of
the callosity pattern. The process generate similarity scores with extracts from photographs already in the catalogue (Hiby and Lovell 2001). The process use a threedimensional surface model on the head of right whales, allowing to compensate
differences in the angle from which the photograph was taken (Fig. 16.12). The
orientation and scaling of the model to the whale in the photograph is made by
identifying four reference points: the front margin of the bonnet, the rear margin of
the coaming and the upper margin of the eyebrow callosities. The next step is to
identify samples of callosity and skin from the digital image, which the program
then uses to generate an extract of the callosity pattern (Fig. 16.12). The operator
can extend or reduce the selected areas, and also manually paint the callosity area.
The act of painting is almost an artistic work, and must be done with the aid of a
graphic tablet. In addition, the operator has to identify which of the callosities lie
along the lip margins so that the entire image is classified into areas of rostral
callosity, lip callosity and skin. Areas with uncertainty can also be highlighted (Hiby
and Lovell 2001). Information on the date of the photograph, if the lip callosities are
present or absence, as well as if the year of birth are required at the end of this step.
The extracts are initially stored in a pending directory and then automatically compared. The digital images from which the extracts with the best similarity scores
were obtained can be displayed next to the image corresponding to the pending file,
to confirm or reject a match (Fig. 16.13). Each file has a unique name identifying the
date when the photo was taken and a temporary number of the whale. In case a
match is confirmed the file receive the whale number in the library, or a new number
if not a match.
The numbering of the files in the catalog consists of adding a letter (corresponding to the country, e.g. A for Argentina, B for Brazil) and a sequential number identifying the whale. If cataloguing a calf, it receives the mother’s number, with the two
last digits of the year of birth added at the end of the mother’s number. For example
B101-03 is the calf of the whale B101, born in 2003. All the information associated
to the whale in the photograph, e.g. date, sighting location, accompanied whales,
presence of calf, is stored in a database associated to the automated system. The
16 Conservation Advances for the Southern Right Whales in Brazil
In parallel, many whale populations have been recovering, including southern
right whales (IWC 2013) and so the number of sightings and consequently individuals photographed in each aerial survey. This demanded for the development of some
automated system which could assist the comparisons. In 1998, Hiby and Lovel
(2001) developed a computer program called “Right Whale Photo-identification
Software”, to aid on the analysis of aerial photographs taken from right whales.
The “Right Whale Photo-identification Software” makes use of the variation in
shape, number and location of callosities on the rostrum and lips, transforming the
three-dimensional shape of the head of right whales in a two-dimensional picture,
where image pixels define the presence of callosities, generating a library of patterns for comparison with new images analyzed. The objective of the software is to
suggest the order in which the existing catalogue be searched for matches to a new
photograph, and to eliminate from the search those photographs in the library that
cannot possibly be potential matches of the new one (Hiby and Lovell 2001).
The method consists in extract from each new photograph a simplified map of
the callosity pattern. The process generate similarity scores with extracts from photographs already in the catalogue (Hiby and Lovell 2001). The process use a threedimensional surface model on the head of right whales, allowing to compensate
differences in the angle from which the photograph was taken (Fig. 16.12). The
orientation and scaling of the model to the whale in the photograph is made by
identifying four reference points: the front margin of the bonnet, the rear margin of
the coaming and the upper margin of the eyebrow callosities. The next step is to
identify samples of callosity and skin from the digital image, which the program
then uses to generate an extract of the callosity pattern (Fig. 16.12). The operator
can extend or reduce the selected areas, and also manually paint the callosity area.
The act of painting is almost an artistic work, and must be done with the aid of a
graphic tablet. In addition, the operator has to identify which of the callosities lie
along the lip margins so that the entire image is classified into areas of rostral
callosity, lip callosity and skin. Areas with uncertainty can also be highlighted (Hiby
and Lovell 2001). Information on the date of the photograph, if the lip callosities are
present or absence, as well as if the year of birth are required at the end of this step.
The extracts are initially stored in a pending directory and then automatically compared. The digital images from which the extracts with the best similarity scores
were obtained can be displayed next to the image corresponding to the pending file,
to confirm or reject a match (Fig. 16.13). Each file has a unique name identifying the
date when the photo was taken and a temporary number of the whale. In case a
match is confirmed the file receive the whale number in the library, or a new number
if not a match.
The numbering of the files in the catalog consists of adding a letter (corresponding to the country, e.g. A for Argentina, B for Brazil) and a sequential number identifying the whale. If cataloguing a calf, it receives the mother’s number, with the two
last digits of the year of birth added at the end of the mother’s number. For example
B101-03 is the calf of the whale B101, born in 2003. All the information associated
to the whale in the photograph, e.g. date, sighting location, accompanied whales,
presence of calf, is stored in a database associated to the automated system. The
16 Conservation Advances for the Southern Right Whales in Brazil