基于深度学习的远程激光图像特征动态跟踪

详细信息    查看全文 | 推荐本文 |

英文篇名：Dynamic tracking of remote laser image features based on deep learning 作者：周晓梅 ; 卜言彬 英文作者：ZHOU Xiaomei;BU Yanbin;China Communication University; 关键词：远程激光图像 ; 卷积神经网络 ; 深度学习 ; 动态跟踪 ; Harris角点检测 英文关键词：remote laser image;;convolution neural network;;deep learning;;dynamic tracking;;Harris corner detection 中文刊名：JGZZ 英文刊名：Laser Journal 机构：中国传媒大学南广学院; 出版日期：2019-02-25 出版单位：激光杂志 年：2019 期：v.40;No.257 基金：江苏省高校自然科学研究面上项目资助(No:18KJD520006) 语种：中文; 页：JGZZ201902022 页数：5 CN：02 ISSN：50-1085/TN 分类号：109-113

摘要

为了提高对远程激光图像的自动识别和检测能力,提出一种基于深度学习卷积神经网络的远程激光图像特征动态跟踪技术。采用动态块分割技术进行远程激光图像的激光三维成像轮廓检测,采用Harris角点检测方法进行远程激光图像的关键特征点定位,在远程激光图像的分块区域内进行图像的激光亮点动态特征检测和外包矩形轮廓区域特征动态跟踪,结合角点分布特性对远程激光图像特征点进行动态特征点定位跟踪。对提取的远程激光图像特征点采用卷积神经网络进行分类,利用深度学习算法进行收敛性控制,实现对远程激光图像特征的动态跟踪识别。仿真结果表明,采用该方法进行远程激光图像特征动态跟踪的准确性较高,具有较好的特征匹配能力。
        In order to improve the ability of automatic recognition and detection of remote laser images,a dynamic tracking technique for remote laser images based on deep learning convolution neural network is proposed. The dynamic block segmentation technique is used to detect the laser 3 D image contour of remote laser image,and the Harris corner detection method is used to locate the key feature points of the remote laser image. The dynamic feature detection of laser light spots and the dynamic tracking of outsourced rectangular contour region are carried out in the segmented region of remote laser image,and the dynamic feature point location of remote laser image is tracked by combining the corner distribution characteristics. The feature points of the remote laser image are classified by convolution neural network,and the convergence control of the deep learning algorithm is carried out to realize the dynamic tracking and recognition of the feature of remote laser image. The simulation results show that the proposed method is more accurate and has better feature matching ability.

引文

[1] MOGHADDAM Z,PICCARDI M. Training initialization of hidden markov models in human action recognition[J].IEEE Transactions on Automation Science&Engineering,2014,11(2):394-408.
    [2] AMOR B B,SU J,SRIVASTAVA A. Action recognition using rate-invariant analysis,of skeletal shape trajectories[J]. IEEE Transactions on Pattern Analysis&Machine Intelligence,2016,38(1):1-13.
    [3]李松林,贾勇,郭勇,等.基于改进Camshift的穿墙雷达运动人体目标成像跟踪算法[J].计算机应用,2018,38(2):528-532.
    [4]周颖玥,臧红彬,赵井坤,等.基于非局部平均滤波的冲击噪声图像恢复算法[J].计算机应用研究,2016,33(11):3489-3494.
    [5]姜婷婷,肖卫东,张翀,等.基于桑基图的时间序列文本可视化方法[J].计算机应用研究,2016,33(9):2683-2687.
    [6] SHI Zhan,XIN Yu,SUN Yu'e,HUANG He. Task allocation mechanism for crowdsourcing system based on reliability of users[J]. Journal of Computer Applications,2017,37(9):2449-2453.
    [7] RUI L L,ZHANG P,HUANG H Q,et al. Reputationbased incentive mechanisms in crowdsourcing[J]. Journal of Electronics&Information Technology,2016,38(7):1808-1815.
    [8] ZHANG Y,JIANG C,SONG L,et al. Incentive mechanism for mobile crowdsourcing using an optimized tournament model[J]. IEEE Journal on Selected Areas in Communications,2017,35(4):880-892.
    [9]黄鸿,何凯,郑新磊,等.基于深度学习的高光谱图像空-谱联合特征提取[J].激光与光电子学进展,2017,54(10):174-182.
    [10]谢军昱,许杨剑,王效贵.基于贝叶斯模型和数字图像相关的视觉测量[J].激光技术,2016,40(6):866-870.
    [11]唐彦琴,顾国华,钱惟贤,等.四象限探测器基于高斯分布的激光光斑中心定位算法[J].红外与激光工程,2017,46(2):206003-0206003(7).
    [12]牛英宇,聂瑞杰,李丽娟.基于FPGA的红外图像非均匀校正实现方法[J].激光与红外,2016,46(8):1028-1032.
    [13]肖淑苹,贺毅岳.一种改进的EMD图像信号去噪算法[J].现代电子技术,2016,39(16):91-93.
    [14] CARLSON N A,PORTER J R. On the cardinality of Hausdorff spaces and H-closed spaces[J]. Topology&its Applications,2017,160(1):137-142
    [15] ZHANG T,MU D J,REN S,et al. Information hiding scheme for 3D models based on skeleton and inscribed sphere analysis[J]. Journal of Xidian University,2014,41(2):185-190.
    [16] ANG Lifang,CHENG Xi,QIN Pinle,GAO Yuan. Nonrigid multi-modal medical image registration based on multi-channel sparse coding. Journal of Computer Applications,2018,38(4):1127-1133.
    [17]李明爱,崔燕,杨金福,等.基于HHT和CSSD的多域融合自适应脑电特征提取方法[J].电子学报,2013,41(12):2479-2486.