融合渐进式去雨网络的军用车辆检测算法  

Military vehicle detection algorithm combining progressive deraining network

作　　者：苏胜君 仝秋红 柴国庆 苏海东 王凯 胡待方 SU Shengjun;TONG Qiuhong;CHAI Guoqing;SU Haidong;WANG Kai;HU Daifang(School of Automobile,Chang’an University,Xi’an 710018,China;Xi’an Institute of Optics and Precision Mechanics,Chinese Academy of Sciences,Xi’an 710119,China;Shaanxi Intelligent Connected Vehicle Research Institute Co.,Ltd.,Xi’an 710000,China;School of Information Engineering,Chang’an University,Xi’an 710064,China)

机构地区：[1]长安大学汽车学院,陕西西安710018 [2]中国科学院西安光学精密机械研究所,陕西西安710119 [3]陕西智能网联汽车研究院有限公司,陕西西安710000 [4]长安大学信息工程学院,陕西西安710064

出　　处：《现代电子技术》2025年第5期127-134,共8页Modern Electronics Technique

基　　金：国家重点研发计划(2022YFC3002602);“两链”融合企业(院所)联合重点专项-工业领域(2022LL-JB-03)。

摘　　要：针对雨天场景下检测军用车辆目标时出现的精度退化问题,提出一种将渐进式去雨算法与高精确率检测器相融合的军用车辆检测方法。首先设计了一个图像去雨算法HISPNet,其包括轻量级高效雨纹特征提取模块和跨子网雨纹特征融合模块,捕获雨纹信息的同时缓解卷积过程中的细节特征丢失问题;其次引入SPPFCSPC模块改进了单阶段检测器,保证检测器感受野的同时提高了效率,增强了检测模型的表达能力。自建数据集中的实验结果表明,雨天场景下,相较于经典检测算法YOLOv7,所提算法的mAP@0.5、mAP@0.5:0.95分别提升了4.4%、2.8%,算法检测速度达到21.05 f/s,基本满足检测实时性要求,证明了所提算法的有效性与实用性。In view of the accuracy degradation when detecting military vehicles(the objects)in the scenarios of rainy days,a military vehicle detection method that fuses a progressive rain removal algorithm with a high-accuracy detector is proposed.An image deraining algorithm named HISPNet is designed.The algorithm includes a lightweight and efficient rain streak feature extraction module and a cross-subnet rain streak feature fusion module,which mitigates the loss of detailed features during the process of convolution while capturing the rain streak information.The SPPFCSPC(spatial pyramid pooling and fully-connected spatial pyramid convolution)module is introduced to improve the single-stage detector,which ensures the detector sensing field and improves the efficiency at the same time,and enhances the representation ability of the detection model.The experimental results of the self-built dataset show that,in comparison with the classical detection algorithm YOLOv7,the mAP@0.5 and mAP@0.5:0.95 of the proposed algorithm have been improved by 4.4%and 2.8%,respectively,in the scenarios of rainy days,and its detection speed is 21.05 f/s.It can be seen that the proposed algorithm can basically meet the real-time requirements of detection,and is of effectiveness and practicality.

关 键 词：图像去雨 编码器-解码器架构 轻量级高效雨纹特征提取模块 跨子网雨纹特征融合模块 SPPFCSPC模块 军用车辆检测 

分 类 号：TN911.7-34[电子电信—通信与信息系统] TP391.4[电子电信—信息与通信工程]