摘要
针对嵌入式平台性能较低、内存资源少,限制了性能优秀的目标跟踪法的应用,因此文中提出了一种基于ARM平台Android系统的TLD目标跟踪算法的优化和移植方法。为了改进TLD算法的实时性能,提出了基于Camshift跟踪改进跟踪器实时性能,基于背景差法提升检测器前景目标检测效率的方法。通过搭建基于Android系统的DM3730嵌入式硬件平台,利用开源Open CVfor Andorid库和Android NDK方法对改进的TLD算法进行移植。实验表明,改进后的TLD算法在嵌入式平台保持了较好的准确性和实时性。
The poor performance and memory resources of the embedded platform limit the application of highperformance tracking algorithms. This paper proposes a method of optimizing and porting a TLD target tracking algorithm based on ARM platform Android system in order to improve real-time performance of the TLD algorithm. A method based on the Camshift tracking and background-difference method is proposed to promote performance of tracker and detector of TLD in the embedded platform. A DM3730 embedded hardware platform based on Android is built for transplanting the optimized TLD using the NDK OpenCVforAndroid open source libraries. Experimental resuits show that the optimized TLD algorithm achieves good accuracy and real-time performance in the embedded platform.
出处
《电子科技》
2016年第12期130-133,141,共5页
Electronic Science and Technology
