基于卷积神经网络的移动网络目标跟踪与控制  

Mobile network target tracking and control based on convolutional neural network

作　　者：魏冬梅 WEI Dong-mei(School of Computer and Software Engineering,Xihua University,Chengdu 610039,China;School of Management Science and Engineering,Southwestern University of Finance and Economics,Chengdu 610074,China)

机构地区：[1]西华大学计算机与软件工程学院,成都610039 [2]西南财经大学管理科学与工程学院,成都610074

出　　处：《沈阳工业大学学报》2023年第3期342-347,共6页Journal of Shenyang University of Technology

基　　金：教育部2022年第一批产学合作协同育人项目(220606627135356);全国高等院校计算机基础研究会项目(2022-AFCEC-559);西华大学教改项目(xjjg2021060)。

摘　　要：针对移动网络目标跟踪与控制过程中易出现遮挡、形变和模糊等问题,提出了基于卷积神经网络的移动网络目标跟踪与控制方法.利用卷积神经网络构建CNN模型,在原始移动网络目标图像金字塔内,提取多尺度特征,结合滤波算法获取的移动目标响应图,在多模板条件下求解各层卷积特征滤波最佳参数,通过最佳参数和滤波响应获取最终响应图,根据最终响应图中响应最大值实现移动网络目标的自适应跟踪与控制.实验结果表明,所提方法平均跟踪精确度为97.18%,平均跟踪成功率为95.93%,中心位置平均误差为4.38,误差幅值较小,具有良好的目标跟踪与控制能力.In order to solve the problems of occlusion,deformation and fuzzy image in the tracking and controlling process of mobile network,a new method of target tracking and control based on convolutional neural network was proposed.CNN model was constructed by convolutional neural network.In the pyramid of original mobile network target image,multiscale features were extracted,and the response graph of mobile target obtained by filtering algorithm was constructed.The optimal parameters of convolution feature filtering were solved under the condition of multi-templates,and the final response graph was obtained by the best parameters and filtering response.According to the maximum response value in the final response graph,the adaptive tracking and control of mobile network targets was realized.Test results show that the average tracking accuracy of proposed method is 97.18%,the average tracking success rate is 95.93%,the average error of center position is 4.38,and the error amplitude is small,showing good capabilities for target tracking and control.

关 键 词：卷积神经网络 移动网络 目标跟踪 滤波算法 特征提取 响应值 拉普拉斯滤波 高斯金字塔 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]