【摘要】 针对近年来养蜂人数增长，蜜蜂数量减少，导致蜂农经济收入下滑的问题，文章基于EfficientNet深度学习模型，结合深度学习多目标跟踪策略检测(deep simple online and realtime tracking,DeepSORT)算法，构建了一种蜜蜂识别跟踪计数系统。实验表明，该系统在识别准确率、跟踪准确性、计数准确性等多目标跟踪性能方面相较于传统方法呈现出明显的优势，为蜜蜂养殖与生态研究提供了技术支持。更多还原

【Abstract】 Addressing the issue of declining bee populations amidst growing number of beekeepers in recent years—resulting in reduced economic returns for beekeepers,this paper constructs a bee identification,tracking and counting system. The system integrates the EfficientNet deep learning model with the deep simple online and realtime tracking( DeepSORT) algorithm,specifically designed for multi-object tracking tasks. The experimental results show that this system exhibits significant advantages over traditional methods in multi-objective tracking performance, including identification accuracy, tracking precision, and counting reliability. This system provides robust technical support for bee-keeping practices and ecological research.更多还原