– by L. Grøva (NIBIO)
Master student Idun Syvertsen submitted her thesis in Computer Science at NTNU in June 2025 (Norwegian University of Science and Technology (NTNU), Faculty of Information Technology and Electrical Engineering, Department of Computer Science). Title of thesis: “Using deep learning on GPS trajectories to identify abnormal behaviour in sheep//Classifying abnormal behaviour in sheep using CNNs and GPS data//Identifying distress in sheep using GPS data and machine learning”. The main idea behind this work was to investigate if and how GPS data could be used to identify distress in sheep. Position data from GPS collars (here Telespor) were linked with individual sheep data from The Norwegian Sheep Recording System (NSRS) (i.e. Sauekontrollen). The dataset consisted of 62 sick sheep and 383 not sick sheep; this being fewer data than anticipated due to a number of data mining and linking challenges. Distance walked and polygon area used was calculated for healthy and sick ewes. Further, machine learning was considered, and the following tools were selected; K Nearest Neighbors (KNN), Support Vector Machines (SVM), and Random Forest. Also training a CNN-model (Convolutional Neural Network) was tested. Below is a scatter plot showing the distribution of the final labelled data against polygon area and daily average distance for each ewe. Although, data were too few to be conclusive we think there is potential for these data to detect abnormal behavior in sheep.
© all images NIBIO
