基于MC-CNN网络的磁性运动舰船目标分类检测  

Classification and detection of magnetic moving ship targets based on MC-CNN

作　　者：马剑飞 颜冰[1] 林春生[1] 陈浩[1] MA Jian-fei;YAN Bing;LIN Chun-sheng;CHEN Hao(College of Weaponry Engineering, Naval Univ. of Engineering, Wuhan 430033, China)

机构地区：[1]海军工程大学兵器工程学院,武汉430033

出　　处：《海军工程大学学报》2020年第5期67-71,77,共6页Journal of Naval University of Engineering

基　　金：国家部委基金资助项目(41419010208)。

摘　　要：为了提升水下平台对舰船目标的分类检测能力,在椭球体与磁偶极子阵列混合模型的基础上,建立了磁性舰船运动目标三分量投影模型,并据此生成了磁性舰船运动目标在变参数情况下的10类目标训练数据库。进一步提出了MC-CNN网络分类检测算法,实测和仿真磁数据库的分类检测结果表明,该网络具有运算量小、准确率高的特点。另外,鉴于MC-CNN分类准确率随着训练测试深度差的增加而性能变差的缺陷,提出了一种具有小数据测量、分类检测准确特点的工程化应用方式。In order to promote the classification and detection ability of underwater platform for ship targets,a three-axis projection model of magnetic moving ship targets was established based on the hybrid model of ellipsoid and magnetic dipole array,and 10 kinds of target training databases of magnetic moving ship targets with variable parameters were generated.Then the MC-CNN classification algorithm was proposed.The classification results of the measured and simulated magnetic databases show that the algorithm is characterised by less computation with higher accuracy.Furthermore,in view of the disadvantage that the classification accuracy of MC-CNN deteriorates with the difference of training and testing depth increasing,a practical method with the characteristics of small measurement data and accurate classification was proposed.

关 键 词：舰船磁场 投影模型 磁数据库 MC-CNN网络 

分 类 号：TJ-610[兵器科学与技术]