| 摘 要: 针对目前智能摄像头组合应用技术在视频传输质量控制、资源调度与分配、算法智能性与优化等方面存在不足,设计了基于深度学习的物联网智能摄像头控制系统。系统通过深度学习和边缘计算能力在本地设备上分析和预测网络性能需求变化,动态切换最优的相机参数设置,通过强化学习算法制定最佳的协作策略,同步动态调度和分配多台智能摄像头共享的网络资源。系统将检测精度(mAP@0.5)提升至84.7%,协作任务成功率提高23.6%,同时降低平均响应延迟至90ms以下实现智能摄像头组合之间的视频数据传输、质量控制和网络资源分配等多个方面的智能化和精细化控制与管理。 |
| 关键词: 物联网 智能摄像头 控制系统 边缘计算 强化学习 深度学习 |
|
中图分类号: TP181
文献标识码: A
|
| 基金项目: 2024年辽宁省教育厅高校基本科研项目(LJ212410845002) |
|
| Design of an Io T Smart Camera Control System Based on Deep Learning |
|
LIU Jianying
|
(School of Electronics and Information Engineering, Dalian Vocational & Technical College, Dalian 116035, China)
s-il-l y@163.com
|
| Abstract: In response to the current shortcomings in the combined application technology of smart cameras, such as video transmission quality control, resource scheduling and allocation, and the intelligence and optimization of algorithms, a deep learning-based IoT smart camera control system is designed. The system utilizes deep learning and edge computing capabilities to analyze and predict changes in network performance requirements on local devices,dynamically switches to optimal camera parameter settings, and formulates the best collaboration strategy through a reinforcement learning algorithm. It synchronously and dynamically schedules and allocates network resources shared by multiple smart cameras. The system improves detection (mAP @ 0. 5) to 84. 7% , increases the success rate of collaborative tasks by 23.6% , and reduces the average response latency to below 90 ms. It achieves intelligent and refined control and management in various aspects, including video data transmission, quality control, and network resource allocation among combinations of smart cameras. |
| Keywords: Internet of Things smart camera control system edge computing reinforcement learning deep learning |
