面向航拍交通目标的实时检测算法  

Real-time detection algorithm for aerial traffic targets

作　　者：黄林辉 钟小勇[1] 杨浩 邱昊 HUANG Lin-hui;ZHONG Xiao-yong;YANG Hao;QIU Hao(College of Science,Jiangxi University of Science and Technology,Ganzhou 341000,China)

机构地区：[1]江西理工大学理学院,江西赣州341000

出　　处：《计算机工程与设计》2025年第2期587-594,共8页Computer Engineering and Design

基　　金：国家自然科学基金项目(51665019);江西省研究生创新专项资金基金项目(YC2021-S601);江西理工大学研究生创新计划基金项目(XY2022-S200)。

摘　　要：针对目前无人机视角下进行交通目标检测过程中的各项问题,提出一种面向航拍交通目标的实时检测算法。引入高效通道注意力机制,加强网络对重要特征信息的提取能力,设计一种多元特征融合模块,帮助特征融合网络更好掌握全局信息;在此基础上,引入自适应空间特征融合模块,对浅层特征和深层的语义信息进行融合;采用更优的回归损失函数对网络进行训练,获取预测框和真实框之间更精准的位置信息。实验结果表明,该算法比主流算法具有更高的检测精度和检测速度。Aiming at the problems in the process of traffic target detection from the perspective of UAVs,a real-time detection algorithm for aerial traffic targets was proposed.An efficient channel attention mechanism was introduced to enhance the ability of the network to extract important feature information,and a multi-feature fusion module was designed to help the feature fusion network grasp the global information better.On this basis,an adaptive spatial feature fusion module was introduced to integrate shallow features and deep semantic information.A better regression loss function was used to train the network to obtain more accurate position information between the predicted frame and the real frame.Experimental results show that the proposed algorithm has higher detection accuracy and speed than the mainstream algorithm.

关 键 词：无人机 交通目标检测 注意力机制 多元特征融合 自适应空间特征融合 小目标检测 损失函数 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]