计量学报2025，Vol.46Issue(10)：1475-1485,11.DOI:10.3969/j.issn.1000-1158.2025.10.10
基于YOLOv8-seg的近场MIMO-SAR危险目标检测方法研究
Research on Dangerous Target Detection Method of Near-field MIMO-SAR Based on YOLOv8-seg
摘要
Abstract
An image target detection algorithm of near-field multiple input and multiple output synthetic aperture radar(MIMO-SAR)millimeter-wave is proposed based on YOLOv8-seg to resolve the problems of missed detection,false detection,and low precision in detecting near-field SAR images with existing target detection methods.This algorithm is designed for dangerous target detection in near-field scenarios.The CBS(Convolutional-Batchnormal-SiLu)of the Backbone and the Neck of YOLOv8-seg is replaced by GhostConv to reduce the algorithm parameters and realize the lightweight algorithm.Moreover,the C3-RVB module is constructed by replacing the Bottleneck in CSPDarknet53 by RepViT Block,and the C2f module in the Neck of YOLOv8-seg is replaced with C3-RVB to increase feature extraction capability and improve the detection accuracy.Furthermore,the loss function CIoU of YOLOv8-seg is superseded by Inner-EIoU,and a scaling factor is introduced to control the scale of the auxiliary bounding box to enhance the generalization ability of the algorithm.MIMO-SAR dataset containing five types of dangerous targets is built.Experimental results demonstrate that the improved YOLOv8-seg algorithm achieves high detection accuracy with 97.3%mAP@0.5 and 2.56M parameters.The issues of missed and false detection of conventional YOLOv8-seg are resolved effectively.Additionally,the algorithm is lightweight with fewer parameters and offers fast detection speed.The 3DRIED dataset further validates that the improved algorithm possesses robust generalization capabilities.关键词
危险目标检测/YOLOv8n-seg/MIMO-SAR成像/图像处理/毫米波雷达Key words
dangerous target detection/YOLOv8n-seg/MIMO-SAR imaging/image processing/millimeter wave radar引用本文复制引用
林纪闳,张远辉,刘康,张华峰,李运堂..基于YOLOv8-seg的近场MIMO-SAR危险目标检测方法研究[J].计量学报,2025,46(10):1475-1485,11.基金项目
浙江省自然科学基金(LY19F010007) （LY19F010007）
