基于深度学习的三维目标识别技术现状  

Status Quo of 3D Target Recognition Technology Based on Deep Learning

作　　者：柳碧辉 王培元 LIU Bihui;WANG Peiyuan(Naval Aviation University,Yantai 264001,China)

机构地区：[1]海军航空大学,山东烟台264001

出　　处：《兵器装备工程学报》2021年第8期140-146,共7页Journal of Ordnance Equipment Engineering

摘　　要：以图像目标识别、三维目标识别和深度学习的基本概念为切入点,论述了深度学习在目标识别和三维识别中的研究现状,分析了现阶段基于深度学习的三维识别方法所存在的不足和研究方向,进一步探讨了三维目标识别技术的应用前景。通过对相关技术和文献的调研和分析,对多种基于深度学习的目标识别优化结构及优化算法的优缺点进行了总结,说明了三维目标识别技术取得的进步和对社会生活、工业生产等领域产生的深远影响;指出典型目标全方位识别、特定目标要害部位打击、战场态势感知等方面的应用是三维识别技术未来研究的重点,且需进一步关注基于深度学习的算法样本库的更新,增强识别的实时性和真实性。The basic concepts of image target recognition,3D target recognition and deep learning were taken as the starting point,and the research status of deep learning in target recognition and 3D recognition was discussed.We analyzed the current deficiencies and research directions of 3D recognition methods based on deep learning,and further discussed the application prospects of 3D target recognition technology.Through the investigation and analysis of related technologies and literature,the advantages and disadvantages of a variety of deep learning-based target recognition optimization structures and optimization algorithms were summarized,and the progress of the three-dimensional target recognition technology and its impact on social life,industrial production and other fields were summarized.The far-reaching impact pointed out that the application of omni-directional recognition of typical targets attack on key parts of specific targets.The battlefield situational awareness is still the focus of future research on 3D recognition technology,and further attention needs to be paid to the update of the algorithm sample database based on deep learning to enhance the real-time and authenticity of recognition.

关 键 词：目标识别 三维识别 深度学习 三维模型 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]