基于鲁棒MPC的移动视觉目标检测性能优化  

Performance Optimization of Object Detection in Mobile Vision Based on Robust MPC

作　　者：李昌镐 谭光 LI Chang-hao;TAN Guang(School of Intelligent Systems Engineering,Sun Yat-sen University,Shenzhen Guangdong 518106,China)

机构地区：[1]中山大学智能工程学院,广东深圳518106

出　　处：《计算机仿真》2022年第5期191-195,200,共6页Computer Simulation

基　　金：国家自然科学基金资助项目(61772509);广东省自然科学基金资助项目(2019A1515011066)。

摘　　要：随着终端智能化的快速提高,AR、VR等移动端设备大量涌现,然而移动端目标检测任务经常收到设备计算能力和能量储备的限制,无法满足用户对检测任务体验质量的要求。针对以上问题,提出了一个端云协同的移动视觉检测系统,在此系统中利用鲁棒模型预测控制算法对目标检测与目标跟踪进行决策调度,以提高用户体验质量。通过实验证明,上述调度算法能够在复杂场景变化条件下保证一定的检测精度,达到1.9%-67.5%的用户体验质量提升和3.3%-51.5%的能耗降低,使移动视觉目标检测系统达到最优性能。With the rapid improvement of terminal intelligence,a large number of mobile terminal devices,such as AR and VR,have emerged.However,mobile devices are often limited by the computing power and energy budgets,and cannot meet users’ requirements for the quality of experience on video detection tasks.In order to improve the quality of user experience,an object detection system in mobile vision based on device-cloud collaboration is proposed in this paper.In this system,the robust model predictive control algorithm is used to make decision scheduling for object detection and object tracking.The results show that the proposed scheduling algorithm can guarantee certain detection accuracy under complex scene changes,improve the quality of user experience by 1.9%-67.5% and reduce the energy consumption by 3.3%-51.5%,so that the object detection system in mobile vision can achieve the optimal performance.

关 键 词：端云协同 体验质量 目标检测与跟踪 决策调度 

分 类 号：TP393.09[自动化与计算机技术—计算机应用技术]