Classification of flying bats using computer vision techniques

We are developing computer vision techniques to automatically monitor bat populations, and extract biometric features which willbe used to gather important population data. The biometric features will include shape, speed, trajectory features, and wing beatfrequency. We will then use classifiers built using Support Vector Machines (SVM) and Neural Networks, to classify bats intospecies type, male, female, pregnant and young by tracking individual bats in 2D and 3D in low-light using standard camerasThe Department for environment, food and rural affairs (DEFRA) in association with the Bat Conservation Trust (BCT) started anational bat monitoring programme in 1996. Questions that their surveys seek to answer include: Which species are affected byhabitat changes? What are bats’ hibernation habits? And how many bats at roosting site are females/males, young, pregnant etc.?Bat populations also roost in buildings, including historic buildings such as churches. This habitation often leads to damage tobuilding fabric and sensitive artefacts. Data about these populations enables the effective management and protection of thebuildings they inhabit, and we anticipate that our work will be useful not only to conservationist studying bats, but also to buildingmanagers and professional ecologists surveying these buildings.