Automatic localisation and behaviour detection of individual animals through video image processing

PhD type sandwich thesis between UK and Vietnam
Laboratory 1 Laboratory 2

MICA Institute

Hanoi University of Science and Technology

1 Dai Co Viet str.

Hanoi

Vietnam

University of Strathclyde in Glasgow

16 Richmond Street,

Glasgow G1 1XQ,

United Kingdom

 

 

Supervisors

Dr Dao Trung Kien - trung-kien.dao at mica.edu.vn

Dr Le Thi Lan - thi-lan.le at mica.edu.vn

 

Dr Christos Tachtatzis

Dr Paul Murray

Subject

Agricultural resource management is critical both for the Vietnamese and UK economy. The average UK farm size has almost doubled in the last decade due to increased food demand, and similar trends are expected in Vietnam. Farmers now are becoming more reliant on technology to help observe their herd. This can be seen in the increasing adoption of technology through oestrus or, ‘heat’, detection collars, pedometers and microphone used to optimise fertility and animal welfare. Farmers and animal scientists are particularly interested in animal activity/behaviour and posture. In particular location, walking, standing, laying down patterns can infer wellbeing and currently this only recorded for research purposes rather than routinely.

Scientists use video footage to record and analyse animal behaviour through the construction of ethograms. This is a labour intensive task and requires operators to watch through the footage, a labour intensive task. It is estimated that for every hour of footage, two man-hours are necessary to process and record animal behaviour. Ethograms record well and loosely defined behaviours with the later introducing inter-observer reliability issues.

The aim of this PhD project proposal is to develop a robust and reliable animal-tracking platform that enables automatic localisation and behaviour detection of individual animals through video image processing. The developed algorithms must be able to localise animals in a shed and detect their posture standing, laying down, walking (along with distance walked), eating, drinking and etc. The detected behaviours should be classified and automatically construct ethograms for further processing and inference.

Skills required

Good knowlede in computer vision, image processing, Programming in C++

Biblio

[1] K.H. Zhang, L. Zhang, and M.H. Yang. “Real-time compressive tracking,” European Conference on Computer Vision, pp. 864-877, 2012.

[2] A. Adam, E., Rivlin, I., Shimshoni, “Robust fragments-based tracking using the integral histogram,” IEEE Conference on Com-puter Vision and Pattern Recognition, pp. 798-805, 2006.

[3] B. Babenko, M.H., Yang, S., Belongie, “Robust object tracking with online multiple instance learning,” IEEE Transactions on Pat-tern Analysis & Machine Intelligence, vol. 33, pp. 1619 -1632, 2011.