ASABE Technical Library - Abstract

Click on “Download PDF” for the PDF version or on the title for the HTML version.
If you are not an ASABE member or if your employer has not arranged for access to the full-text, view access and subscription options

Automatic Monitoring of Turkeys: A Vision-Based Approach to Detect and Analyse the Behaviour of Turkeys in Transport Cages Based on Ellipse Fitting

Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan www.asabe.org

Citation:  7th World Congress on Computers in Agriculture Conference Proceedings, 22-24 June 2009, Reno, Nevada (electronic only)  711P0409e.(doi:10.13031/2013.29108)
Authors:   Ahmad Poursaberi, Anette Wichman, Claudia Bahr, Laura Hänninen, Matti Pastell, Daniel Berckmans
Keywords:   Animal welfare, image processing, real-time monitoring, Turkeys, Animal behavior.

The aim in this project was to develop a tool which can reliably automatically score different behaviours performed by turkeys confided in cages of different heights. These behaviors may, after some further evaluation, be used as an assessment of welfare during transport. Eighteen weeks old male turkeys in two different cages with heights of 40 cm (low cage) and 90 cm (high cage) were followed. In the low cage the height was too low for the bird to assume a standing position whereas in the high cage the bird could stand in a normal position and had more opportunities to perform different behaviours like standing and turning. A camera captured videos simultaneously of one turkey in a low and high cage respectively. The four behaviours turning, lying, standing and wing flapping were chosen to represent different behaviours for welfare assessment. An algorithm based on image analysis was developed to classify the behaviour in real time and automatically. A coarse estimation of the bird was obtained through its colour filtering and shape properties. By using image enhancement and binarization techniques, a binary image is produced from each cage. Morphological operators were used to conclude if the cage was empty or not. Then, according to the binary image the boundary of the bird and the parameters of the best fitted ellipse are extracted. To speed up the procedure of finding the turkey the next time with the previous ellipse parameters, the search area was limited in order to update the new position. According to the ellipse parameters in each frame the four behaviours were categorized. Since the distinction between lying and standing postures from single top-view image is difficult and not accurate with ellipse fitting techniques further solutions are tested based on area and functional shape descriptor of the bird. In comparison with the manually labelled reference frame by frame, the algorithm applied on 60 hours video resulted in 100% correct classification of turning, 94% wing flapping, 83% of lying and 71% of standing behaviors.

(Download PDF)    (Export to EndNotes)