Application of machine vision technology to identification of oestrous in the Julia Creek dunnart (Sminthopsis douglasi)

Abstract

Identification of oestrous is an important step in the captive breeding of the Julia Creek dunnart (Sminthopsis douglasi) at David Fleay Wildlife Park (Qld). This is necessary in order to prevent injury to the male and/or the female of this endangered carnivorous marsupial. Oestrous in S. douglasi is currently identified by the presence of cornified cells in the urine. However, oestrous has recently been identified in S. douglasi by behavioural observations. This research aims to identify and quantify reproductive behaviour through the application of machine vision technology, thus reducing the amount of human observation required in the identification of oestrous in this and similar species. Images from two adjoining cages containing a male and a female S. douglasi were captured on video via a CCD camera. The cages were illuminated by an array of infrared light-emitting diodes. Contrast between the animals and the background in successive frames of the video stream enables the location of the dunnart in each image to be defined. This location is recorded in terms of vertical and horizontal position based on a local coordinate system. This stream of coordinate data can be transformed into behavioural information. Behaviours useful in identifying oestrous have been validated in a previous study.