基于MSER的车载毫米波雷达SAR图像目标检测  

MSER for Automotive Millimeter Wave SAR Image Target Detection

作　　者：王佳慧 杨果 王雨晗 毛新华[1] WANG Jiahui;YANG Guo;WANG Yuhan;MAO Xinhua(College of Electronic and Information Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing Jiangsu 211106,China)

机构地区：[1]南京航空航天大学电子信息工程学院,江苏南京211106

出　　处：《现代雷达》2025年第1期1-7,共7页Modern Radar

基　　金：国家自然科学基金资助项目(62071225);航空科学基金资助项目(2020Z017052001);江苏省自然科学基金资助项目(BK20211181)。

摘　　要：为了推进辅助泊车技术,需要向驾驶员提供准确且可视化的行车场景信息。由于车载环境的复杂性以及车载毫米波雷达合成孔径雷达(SAR)图像中干扰信息的杂多,难以直接通过目标分离实现目标检测。文中提出一种基于最大稳定极值区域(MSER)的车载毫米波雷达SAR图像目标检测方法。此方法基于车辆成像后的尺寸的先验信息,通过检测车载毫米波雷达SAR图像MSER区域、划分并标记空闲停车区域向驾驶员提供车位使用情况,利用形态学滤波平滑MSER区域边界,优化目标检测结果。同时,根据检测结果以自适应宽度的矩形框标记目标保证目标检测的准确性。最后通过对实测数据的仿真与分析验证了所提方法的有效性。To promote the assisted parking(AP)technology,accurate and visual actual scene information needs to be provided to divers.Due to the complexity of the automotive environment and the variety of interference information in the automotive millimeter wave(MMW)radar SAR images,it is difficult to achieve target detection by target separation directly.Therefore,a target detection method is proposed for automotive MMW radar SAR image based on MSER.This method is based on a priori information of the size of the automotive after imaging,and provides drivers with parking space usage by detecting the MSER of the SAR image,dividing and marking vacant parking area.Morphological filtering is used to smooth the boundary of MSER regions to optimize the target detection results.At the same time,the accuracy of target detection is endured by marking the target with an adaptive width rectangle box.Finally,the effectiveness of the proposed method is verified by simulation and analysis of the measured data.

关 键 词：车载毫米波雷达 合成孔径雷达 目标检测 最大稳定极值区域 形态学滤波 

分 类 号：TN957.5[电子电信—信号与信息处理]