基于卷积神经网络的运动模糊编码点识别  

Recognition of Motion Blurred Coded Targets Based on Convolutional Neural Network

作　　者：陈明军 周含策[2] 张丽艳[1] Chen Mingjun;Zhou Hance;Zhang Liyan(College of Mechanical and Electrical Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing 210016;College of Science, Nanjing University of Aeronautics and Astronautics, Nanjing 211106)

机构地区：[1]南京航空航天大学机电学院,南京210016 [2]南京航空航天大学理学院,南京211106

出　　处：《计算机辅助设计与图形学学报》2017年第10期1844-1852,共9页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金(51575276)

摘　　要：为解决运动目标三维视觉测量中的运动模糊视觉特征难以识别的问题,提出一种利用卷积神经网络识别布设于运动目标上的、具有一定运动模糊效应的视觉编码点的方法.首先构建并解析运动模糊编码点识别网络(MBCNet);然后通过分析运动模糊效应形成机理,设计实现六参数驱动的运动模糊图像模拟生成系统,并利用该系统模拟生成的100类编码点,共对66.5万幅运动模糊图像进行网络训练和测试,以解决大量实拍样本数据难以获得的问题.对实际拍摄的5类编码点共1.5万幅的运动模糊图像进行实验的结果表明,其识别精度达到了92.51%;该方法模拟生成的编码点运动模糊图像可以获得良好的网络训练效果,且构建的MBCNet具有良好的泛化性能.In order to recognize the motion blurred coded targets in3D vision measurement of moving objects,a method based on the convolutional neural network was proposed.First,a carefully designed convolutional neural network(MBCNet)was constructed and analyzed.Second,by exploiting the motion blur formation mechanism,a motion blurred image simulating system driven by six parameters was designed,so as to provide the large amount of motion blurred images of the coded targets needed for the network training.Up to665000simulated motion blurred images of100kinds of coded targets were automatically generated and used for training the network.The recognition accuracy of15000real motion blurred images of5kinds of coded targets reached up to92.51%,which shows that the method for simulating motion blurred images is effective and the proposed network MBCNet has good generalization performance.

关 键 词：机器视觉 图像识别 运动模糊 卷积神经网络 编码点 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]