基于无人艇的导航雷达目标检测跟踪算法  

Navigation radar target detection and tracking algorithm based on USV

作　　者：王伟[1] 杜旭洋 杨志伟 吴凡 WANG Wei;DU Xuyang;YANG Zhiwei;WU Fan(School of Intelligent Science and Engineering,Harbin Engineering University,Harbin 150001,China;China Ship Development and Design Center,Wuhan 430064,China)

机构地区：[1]哈尔滨工程大学智能科学与工程学院,黑龙江哈尔滨150001 [2]中国舰船研究设计中心,湖北武汉430064

出　　处：《系统工程与电子技术》2024年第5期1561-1572,共12页Systems Engineering and Electronics

基　　金：国家自然科学基金(62271163);中央高校基础研究基金(3072022QBZ0401,3072021CFT0404)资助课题。

摘　　要：在无人艇利用导航雷达进行环境感知的过程中,针对雷达回波图中出现的区域破碎现象以及对运动目标进行跟踪时存在较大误差的问题,提出了一种应用于导航雷达的目标检测跟踪方法提高无人艇对水面目标的检测能力。首先,对雷达原始回波图像解析并进行预处理操作;其次,在图像连通的基础上,设计自适应阈值分割Hausdorff匹配算法对回波图和地图进行匹配,区分属于目标和陆地的回波;然后,对连续两帧的雷达回波图进行目标匹配;最后,通过加入预测序列模型的经验模态分解算法优化检测跟踪结果,提高获取目标信息的准确性。实验验证结果表明:对1 km内相对运动速度低于30节的水面目标,所提方法目标检测概率提升了6.5%,距离误差低于2%,航速误差低于6%,航向误差低于6°,整体性能优于工程中常用的检测跟踪方法。A target detection and tracking method for navigation radar is proposed to improve the detection ability of unmanned surface vessel for surface targets.This method addresses the problems of fragmented regions in the radar echo map and large errors in tracking moving targets during the process of environmental perception using navigation radar.Firstly,the original echo image of the radar is analyzed and corrected to obtain the required echo image.Secondly,based on image connectivity,a self-adaptive threshold segmentation Hausdorff matching algorithm is designed to match the echo map and the map,distinguishing the echoes belonging to the target and the land.Thirdly,the target matching is performed on the continuous two frames of radar echoes.Finally,the empirical model decomposition(EMD)algorithm of the predictive sequence model is added to optimize the detection and tracking results,and improve the accuracy of obtaining target information.The results of experimental verification show that for surface targets with a relative motion speed of less than 30 knots within 1 km,the distance error is less than 2%,the target detection probability increases by 6.5%,the speed error is less than 6%,and the heading error is less than 6°using this method.The overall performance is better than that of the detection and tracking methods commonly used in engineering.

关 键 词：环境感知 导航雷达 连通算法 地图匹配 经验模态分解算法 

分 类 号：TN953[电子电信—信号与信息处理]