Skip to main content
ARS Home » Plains Area » Las Cruces, New Mexico » Range Management Research » Research » Publications at this Location » Publication #197372
Title: ROBUST CLASSIFICATION OF ANIMAL TRACKING DATA

Author
item SCHWAGER, MAC - MIT
item Anderson, Dean
item BUTLER, ZACK - ROCHESTER INST OF TECH
item RUS, DANIELA - MIT

Submitted to: Computers and Electronics in Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/22/2006
Publication Date: 4/1/2007
Citation: Schwager, M., Anderson, D.M., Butler, Z., Rus, D. 2007. Robust classification of animal tracking data. Computers and Electronics in Agriculture. 56:46-59.

Interpretive Summary: One of the major challenges facing autonomous electronic data acquisition from free-ranging animals is determining an adequate rate of data capture of biotic data while using abiotic factors such as power and storage requirements in an optimum fashion. Data capture rate was evaluated a priori using a computer classification program capable of distinguishing periods of activity from inactivity from cow location and head position data collected approximately every minute. Results of the classification were repeatable among animals and over different time periods. In general, during periods of activity frequent data acquisition appears necessary while during inactivity infrequent sampling appears adequate, this difference in data capture rate can translate directly into power savings when electronic equipment designed to monitor animal behavior is worn by free-ranging animals.

Technical Abstract: This paper describes an application of the K-means classification algorithm to categorize cow tracking data into various classes of behavior. It is found that even without explicit consideration of biological factors, the clustering algorithm can repeatably resolve cow behavior into two groups corresponding to active and inactive periods. Furthermore, it is shown that this classification is robust to a large range of data sampling intervals. An adaptive data sampling algorithm is suggested for improving the energy efficiency and memory usage of animal tracking equipment.