Precision livestock farming goals to individually and mechanically monitor animal exercise to make sure their well being, well-being, and productiveness. Laptop imaginative and prescient has emerged as a promising instrument for this goal. Nevertheless, precisely monitoring people utilizing imaging stays difficult, particularly in group housing the place animals could have related appearances. Shut interplay or crowding amongst animals can result in the loss or swapping of animal IDs, compromising monitoring accuracy.
Goal: To handle this problem, we applied a framework combining a tracking-by-detection technique with a radio frequency identification (RFID) system.
Strategies: We examined this method utilizing twelve pigs in a single pen as an illustrative instance. Three of the pigs had distinctive pure coat markings, enabling their visible identification throughout the group. The remaining pigs both shared related coat shade patterns or have been totally white, making them visually indistinguishable from one another. We employed the newest model of the You Solely Look As soon as (YOLOv8) and BoT-SORT algorithms for detection and monitoring, respectively. YOLOv8 was fine-tuned with a dataset of three,600 pictures to detect and classify completely different pig courses, attaining a imply common precision of all of the courses of 99%. The fine-tuned YOLOv8 mannequin and the tracker BoT-SORT have been then utilized to a 166.7-min video comprising 100,018 frames.
Outcomes: Outcomes confirmed that pigs with distinguishable coat shade markings may very well be tracked 91% of the time on common. For pigs with related coat shade, the RFID system was used to determine particular person animals after they entered the feeding station, and this RFID identification was linked to the picture trajectory of every pig, each . The 2 pigs with related markings may very well be tracked for a mean of 48.6 min, whereas the seven white pigs may very well be tracked for a mean of 59.1 min. In all circumstances, the monitoring time assigned to every pig matched the bottom reality 90% of the time or extra.
Conclusion: Thus, our proposed framework enabled dependable monitoring of group-housed pigs for prolonged intervals, providing a promising various to the unbiased use of picture or RFID approaches alone. This method represents a big step ahead in combining a number of gadgets for animal identification, monitoring, and traceability, significantly when homogeneous animals are saved in teams.
Mora M, Piles M, David I, Rosa GJM. Integrating pc imaginative and prescient algorithms and RFID system for identification and monitoring of group-housed animals: an instance with pigs. Journal of Animal Science. 2024: skae174, https://doi.org/10.1093/jas/skae174
