摘要
针对现有无人机识别准确率不高、难以从视频图像中准确提取无人机图像、难以对无人机分类等问题,利用深度卷积神经网络自行学习图像特征的优势,提出了一种基于深度卷积神经网络的无人机识别方法。首先使用SSD算法对视频图像做无人机目标检测;然后通过训练一个基于VGG16的学习网络,得到一个高效识别模型;将检测到的无人机图片送入VGG16模型进行特征提取;最后完成多种无人机间的分类。在网络模型优化阶段采用了BP算法,提高了识别方法的鲁棒性。实验结果表明,该方法具有较高的识别准确率和良好的工程应用前景。
Deep Convolutional Neural Network(DCNN)can extract the image features automatically.An Unmanned Aerial Ve hicle(UAV)recognition method based on DCNN is proposed to solve the problems of low detection accuracy,difficulty in accurate ly extracting the UAV picture from the video,and classifying of different UAVs.Firstly,the UVA targets are detected from the video by using Single Shot MultiBox Detector(SSD)algorithm.Then an efficient model of recognition is obtained through training a learn ing network based on Visual Geometry Group(VGG)16.The UVA detected images are put into VGG16 model for feature extrac tion.Finally,the classification of different UAVs is accomplished.Back Propagation(BP)algorithm is introduced to improve the ro bustness of the method in network model optimizing phase.Experiments show that the method has higher recognition rate and better engineering application.
作者
刘佳铭
LIU Jiaming(Navy Equipment Department,Beijing 100071)
出处
《舰船电子工程》
2019年第2期22-26,共5页
Ship Electronic Engineering
关键词
深度卷积神经网络
无人机分类
无人机识别
特征提取
识别模型
deep convolutional neural network
UAV classification
UAV recognition
feature extraction
recognition model
