We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Multi-Object Tracking Algorithm for Poultry Behavior Anomaly Detection.
- Authors
Khairunissa, Jasmine; Wahjuni, Sri; Soesanto, Iman Rahayu H.; Wulandari; Akbar, Auriza R.; Rahmawan, Hendra
- Abstract
One of the five freedom principles in animal welfare is freedom from pain, injury, and diseases. Chickens that don't move for a long time are more likely to feel uncomfortable with the environment or with their own body. Observing each poultry manually will waste a lot of time and effort of the farmer, therefore it is necessary to have a way to monitor the poultry. One way that can be done is to develop a chicken movement monitoring system using the Multi-Object Tracking (MOT) algorithm. The MOT algorithm is an object tracking method consisting of object detection and tracking. In previous studies, a hybrid method was used which only detects objects every few frames and tracks objects in that period alternately with the object detection stage. This method produces a Multi-Object Tracking Precision (MOTP) score of 60.4%. Meanwhile, in this study a sequential method is used, where the program performs object detection and object tracking stage sequentially, which results in a MOTP score of 87.64%. Moreover, to detect chickens that have not moved for a long time, the Isolation Forest anomaly algorithm is used. The results of this study can integrate into a real-time chicken coop remote visual monitoring.
- Subjects
OBJECT recognition (Computer vision); TRACKING algorithms; POULTRY; ANIMAL welfare; CHICKEN coops
- Publication
International Journal of Advances in Soft Computing & Its Applications, 2023, Vol 15, Issue 1, p159
- ISSN
2710-1274
- Publication type
Article
- DOI
10.15849/IJASCA.230320.11