To address the issue of honey production reduction caused by wasp infestations in beekeeping, this paper proposes a YOLOv3-K210-based wasp recognition algorithm. Building upon the YOLOv3 algorithm, we optimize it by lightweighting the YOLOv3 architecture and deploying the algorithm model on to the edge computing K210 hardware platform. By offloading the wasp recognition task to edge computing and integrating edge computing with cloud computing technologies, we alleviate network communication pressure and achieve real-time identification of wasp intrusions alongside client-side early warning functionality. Experimental results demonstrate that the trained YOLOv3 model achieves an average recognition rate of 99.14% for four types of wasps, showcasing high accuracy in detection.