现代计算机2024,Vol.30Issue(18) :108-111.DOI:10.3969/j.issn.1007-1423.2024.18.020

基于改进FBMC算法的单目标识别跟踪技术研究

Research on single target recognition and tracking technology based on improved FBMC algorithm

黄惇晶
现代计算机2024,Vol.30Issue(18) :108-111.DOI:10.3969/j.issn.1007-1423.2024.18.020

基于改进FBMC算法的单目标识别跟踪技术研究

Research on single target recognition and tracking technology based on improved FBMC algorithm

黄惇晶1
扫码查看

作者信息

  • 1. 闽西职业技术学院信息与网络中心,龙岩 364000
  • 折叠

摘要

单目标识别跟踪是机器视觉中的主要研究热点之一,许多学者相继提出了目标跟踪算法.设计了一种基于改进FBMC算法的单目标识别跟踪算法(IFBMC),将注意力机制引入到FBMC算法中,在输入的视频图像中增加滤波器来处理相邻图像帧之间的重叠部分,利用注意力机制使得滤波器更加关注目标区域,此外加入了平移、翻转等图像增强技术对样本进行补充.通过在数据的测试集上设计实验,证明了IFBMC算法对于跟踪精度有一定的提升效果.

Abstract

Single target recognition and tracking is one of the main research hotspots in machine vision,many scholars have proposed target tracking algorithms.A Single target recognition and Tracking Algorithm based on Improved FBMC(Filter-bank multicarrier)Algorithm is proposed,a filter is added to the input video image to process the overlapped parts between adjacent im-age frames,and the attention mechanism is used to make the filter pay more attention to the target region,in addition,image en-hancement techniques such as translation and flip are added to supplement the samples.By designing the experiment on the test set of the data,the IFBMC algorithm is proved to be effective in improving the tracking accuracy.

关键词

相关滤波/单目标跟踪/FBMC/深度学习/注意力机制

Key words

correlation filtering/single target tracking/FBMC/deep learning/attention mechanism

引用本文复制引用

出版年

2024
现代计算机
中大控股

现代计算机

影响因子:0.292
ISSN:1007-1423
段落导航相关论文