目标检测是计算机视觉的基石,在医疗成像、工业缺陷检测、自动驾驶和农业监测等领域有着广泛的应用。YOLO(You Only Look Once)作为单阶段目标检测的开创性算法,以其卓越的实时性和高效性成为计算机视觉领域的核心技术。从2016年YOLOv1...目标检测是计算机视觉的基石,在医疗成像、工业缺陷检测、自动驾驶和农业监测等领域有着广泛的应用。YOLO(You Only Look Once)作为单阶段目标检测的开创性算法,以其卓越的实时性和高效性成为计算机视觉领域的核心技术。从2016年YOLOv1的首次亮相到2025年YOLOv12的发布,YOLO通过持续优化网络架构、改进损失函数设计和创新数据增强策略,显著提升了检测性能和应用适配性。本文系统回顾了YOLO长达十年的发展历程,分析了其在不同领域的应用,并探讨了其快速发展背后的驱动力,包括深度学习技术的突破、开源社区的广泛协作以及对实时性和边缘计算的迫切需求。与双阶段算法(如Faster R-CNN)和基于Transformer的方法(如DETR)相比,YOLO在速度和工程部署上展现出显著优势,但在小目标检测、密集场景处理和复杂背景下的鲁棒性方面仍面临挑战。未来YOLO或可通过整合轻量级Transformer模块、优化多任务学习框架以及引入生成式AI增强数据多样性,有望进一步提升精度和场景适应能力。展开更多
The mean shift tracker has difficulty in tracking fast moving targets and suffers from tracking error accumulation problem. To overcome the limitations of the mean shift method, a new approach is proposed by integrati...The mean shift tracker has difficulty in tracking fast moving targets and suffers from tracking error accumulation problem. To overcome the limitations of the mean shift method, a new approach is proposed by integrating the mean shift algorithm and frame-difference methods. The rough position of the moving tar- get is first located by the direct frame-difference algorithm and three-frame-difference algorithm for the immobile camera scenes and mobile camera scenes, respectively. Then, the mean shift algorithm is used to achieve precise tracking of the target. Several tracking experiments show that the proposed method can effectively track first moving targets and overcome the tracking error accumulation problem.展开更多
文摘目标检测是计算机视觉的基石,在医疗成像、工业缺陷检测、自动驾驶和农业监测等领域有着广泛的应用。YOLO(You Only Look Once)作为单阶段目标检测的开创性算法,以其卓越的实时性和高效性成为计算机视觉领域的核心技术。从2016年YOLOv1的首次亮相到2025年YOLOv12的发布,YOLO通过持续优化网络架构、改进损失函数设计和创新数据增强策略,显著提升了检测性能和应用适配性。本文系统回顾了YOLO长达十年的发展历程,分析了其在不同领域的应用,并探讨了其快速发展背后的驱动力,包括深度学习技术的突破、开源社区的广泛协作以及对实时性和边缘计算的迫切需求。与双阶段算法(如Faster R-CNN)和基于Transformer的方法(如DETR)相比,YOLO在速度和工程部署上展现出显著优势,但在小目标检测、密集场景处理和复杂背景下的鲁棒性方面仍面临挑战。未来YOLO或可通过整合轻量级Transformer模块、优化多任务学习框架以及引入生成式AI增强数据多样性,有望进一步提升精度和场景适应能力。
基金supported by the Fundamental Research Funds for the Central Universities Project(CDJZR10170010)
文摘The mean shift tracker has difficulty in tracking fast moving targets and suffers from tracking error accumulation problem. To overcome the limitations of the mean shift method, a new approach is proposed by integrating the mean shift algorithm and frame-difference methods. The rough position of the moving tar- get is first located by the direct frame-difference algorithm and three-frame-difference algorithm for the immobile camera scenes and mobile camera scenes, respectively. Then, the mean shift algorithm is used to achieve precise tracking of the target. Several tracking experiments show that the proposed method can effectively track first moving targets and overcome the tracking error accumulation problem.
