基于电视末制导技术的舰船目标分割与匹配方法在复杂背景下的应用  

Ship segmentation and recognition technology of TV-guided projectiles in complex background

作　　者：张嘉文 史金光[1] 徐东辉 刘佳佳 Zhang Jiawen;Shi Jinguang;Xu Donghui;Liu Jiajia(School of Energy and Power Engineering,Nanjing University of Science and Technology,Nanjing 210094,China;The Second Research Institute of Liaoshen Industries Group Co.,Ltd.,Shenyang 110045,China)

机构地区：[1]南京理工大学能源与动力工程学院,江苏南京210094 [2]辽沈工业集团有限公司研发中心设计二所,辽宁沈阳110045

出　　处：《南京理工大学学报》2021年第5期529-536,共8页Journal of Nanjing University of Science and Technology

基　　金：国防预先研究项目(201710016004)。

摘　　要：针对电视末制导炮弹在复杂背景下对多目标舰船识别精度不高的问题,提出一种结合经典阈值分割和Prewitt边缘梯度算子的方法,对图像中存在的多个舰船轮廓实现清晰的分割,利用模板匹配,选用加速稳健特征(Speed-up robust features,SURF)算法检测特征点,将处理后的舰船图像与弹载图像进行匹配;针对匹配过程中的错误匹配问题,采用M估计采样一致性算法,有效提升了匹配正确率。数值仿真表明,弹上图像处理器在进行单帧图像处理时,通过与传统图像分割技术如坎尼梯度算子、高斯拉普拉斯(Laplacian-of-Gaussian,LoG)梯度算子和最大类间方差法对比,该文提出的算法可以清晰地将视场内的多只舰船进行有效分割,并显著减少了错误匹配。Aiming at the problem that the TV-guided projectiles have low accuracy in identifying multi-target ships under complex backgrounds,a method combining classic threshold segmentation and Prewitt edge gradient operator is proposed to achieve clear segmentation of multiple ships in the image.Using template matching,the speed-up robust features algorithm is used to detect the feature points,and the processed ship image is matched with the missile-borne image.To reduce the problem of incorrect matching in the matching process,a consensus algorithm based on M-estimation is used to improve the accuracy of matching effectively.Numerical simulations show that when the on-board image processor performs single-frame image processing,comparing with the traditional image segmentation techniques such as Canny gradient operator,Laplacian-of-Gaussian(LoG)and maximum inter-class-difference method,the algorithm proposed in this paper can clearly divide the multiple ships in the field of view effectively and reduce the mismatch significantly.

关 键 词：电视末制导炮弹 图像分割 边缘检测 特征点检测 模板匹配 M估计采样一致性算法 

分 类 号：TJ413.6[兵器科学与技术—火炮、自动武器与弹药工程]