Panda facial region detection based on topology modelling

In the surveillance of rare animals, robust region detection is of great importance. Detected animal regions have many important applications in animal biometrics, such as behaviour analysis, gait detection and movement tracking. This paper presents a simple and effective method on facial region detection of pandas, which are the national treasure among animals. Firstly, the black colour of the panda hair is modelled. Secondly, the region growing algorithm is applied to extract connected hair regions. Thirdly, the topology of panda facial region is modelled and constructed. Extensive experiments are carried out to validate the detected facial regions of panda, by specifying the panda´s ears, eyes and nose.