Automatic track recognition of footprints for identifying cryptic species

Area of Science:

• Ecology
• Wildlife Biology
• Computational Biology

Background:

• Track recognition is vital for ecological research and monitoring.
• Traditional manual track identification by experts struggles with morphologically similar species.
• Tracking tunnels offer a cost-effective method for large-scale species indexing.

Purpose of the Study:

• To develop and apply an automated image analysis method for species identification of animal tracks.
• To address the limitations of manual identification, particularly for cryptic small-mammal species.
• To test the method's efficacy in identifying three invasive rat species in New Zealand.

Main Methods:

• Utilizing image analysis for automatic species identification of tracks.
• Applying the method to identify footprints of three invasive rat species in New Zealand.
• Validating the accuracy of the automated system against manual identification.

Main Results:

• Automatic footprint recognition achieved >70% accuracy for individual footprints.
• The method reached >83% accuracy when analyzing tracking cards.
• Successfully detected a recent invasion of a rat species on a previously rat-free island.

Conclusions:

• Automated track recognition provides a reliable and efficient method for species identification.
• The technique significantly improves estimates of species presence and community composition.
• The image analysis approach has broad applicability to various tracked taxa with potential adaptations.