基于Android的自学习视觉跟踪系统设计

Design of Self-Learning Visual Tracking System Based On Android

智能移动设备的快速发展,使其应用到越来越多的方面,目标跟踪便是其中之一。但是,智能移动平台性能较低、内存资源少,限制了目标跟踪算法的应用。因此论文提出了一种TLD目标跟踪算法的优化方法。使其能够在搭载An-droid系统的智能手机上实时运行。通过提出了三种方法提高TLD算法检测模块的效率,提高了TLD算法的实时性。并将改进算法集成到Android应用中,运行在Android智能手机上。实验表明,改进后的TLD算法在主流智能手机上平均帧率在10fps以上,达到实时性要求。并保证了目标跟踪的准确性。

With the rapid development of mobile devices,it has been applied to more and more fields. Object tracking is one of them. However,mobile devices have lower computing power and less memory resources,which limits the application of target tracking algorithms. Therefore,this paper proposes an optimization method for TLD target tracking algorithm. Three methods are proposed to improve the efficiency of TLD algorithm. And the improved algorithm is integrated into the Android application and run on the Android smart phone. Experiments show that the improved TLD algorithm has more than 10 fps while running on the mainstream smart phones,and achieves real-time requirements. And the improved algorithm also guarantees the accuracy of target tracking.