基于机器视觉的非特定目标实时跟踪研究  

Research on real-time tracking of non-specific targets based on machine vision

作　　者：曹超 管沁朴 CAO Chao;GUAN Qinpu(AVIC Electric Measuring Instrument Company,Xi’an 710117,China;Beijing Zhixingzhe Technology Company,102200,China)

机构地区：[1]中航电测仪器股份有限公司,陕西西安710117 [2]北京市智行者科技有限公司,北京102200

出　　处：《电子设计工程》2021年第20期174-179,184,共7页Electronic Design Engineering

摘　　要：计算机视觉在小型无人机及其他便携式设备中的应用越来越广泛,视觉目标跟踪作为其中的关键环节,成为国内外研究的热点问题。现有目标跟踪算法目标单一,应用范围严重受限,无法同时满足跟踪精度和跟踪速度的要求。基于以上背景,文中设计了一种基于相关滤波的多特征融合无人机目标跟踪算法,融合HOG特征和颜色特征作为输入,以最小方差原则在频域内训练出相关滤波器,将颜色特征置信概率叠加到相关滤波响应中,综合选取跟踪结果。通过仿真实验和飞行实验验证了该算法的正确性和有效性。Computer vision is more and more widely used in small drones and other portable devices.As a key link,visual target tracking has become a hot research topic at home and abroad.The existing target tracking algorithm has a single target and severely limited application range,which cannot meet the requirements of tracking accuracy and tracking speed at the same time.Based on the above background,a multi-feature fusion UAV target tracking algorithm is designed based on correlation filtering,which combines HOG features and color features as input,and trains correlation filters in the frequency domain using the principle of minimum variance,and the confidence probability of color features superimposed on the relevant filter response and comprehensively select the tracking result.Simulation experiments and flight experiments verify the correctness and effectiveness of the algorithm.

关 键 词：相关滤波 目标跟踪 多特征融合 机器视觉 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]