| Title |
Real-time Livestock Traceability Number Identification Architecture based on On-device Hybrid Deep Learning in Mobile Edge Environments |
| Keywords |
Edge computing; On-device AI; Livestock traceability; Time-series voting; Optical character recognition |
| Abstract |
This letter proposes an on-device hybrid deep-learning architecture for real-time livestock traceability identification in mobile edge environments. To overcome on-site optical noise and recognition limitations, we constructed a hybrid pipeline integrating a lightweight perception layer, an intelligent reasoning layer based on domain rules and confusion matrix substitution, and a time-series voting algorithm. Experimental results demonstrate the system's practical effectiveness, achieving over 99% identification accuracy and robust real-time responsiveness. |