融合背景信息和方向信息的遥感图像旋转目标检测

打开文本图片集
关键词:遥感图像;旋转目标检测;感受野扩张;方向感知交互注意力中图分类号:TP751 文献标识码:Adoi:10.37188/OPE.20263404.0624 CSTR:32169.14.OPE.20263404.0624
Remote sensing image rotation object detection based on background information and direction information fusion
HUHenan1,MA Aoyang¹,LIRonghua²
(1. Institute ofMechanical Engineering,Dalian Jiaotong University,Dalian 116o28,China; 2. Institute of Electrical Engineering,Dalian Jiaotong University, Dalian 1l6O28, China) * Corresponding author, E -mail: lironghua705@163.com
Abstract:Aimingat the problem of insuficient detection accuracy caused by large differences in target scale and complex direction distribution in remote sensing image target detection,a detection network combining background information and direction information was proposed. Firstly,aimed at the problem of large differences in target scale,a Receptive Field Extending(RFE) module was designed.Different from the traditional fixed receptive field or complex multi-branch structure,this module integrated multi-scale background information through large kernel decomposition,dilated convolution,and paralel branch structure design,and solved the problem of diferent background information requirements for diferent scale targets without significantly increasing the amount of calculation. Secondly,aimed at the problem of complex distribution of target directions,an Orientation Aware Cross Attention (OACA) module was designed.Diffrent from the existing attntion mechanism convolution kernel shape,the module extracted directional texture information through horizontal and vertical separable convolution to prevent feature loss and fracture;at the same time,a cross-atention mechanism was designed to suppress background noise and enhance the interactivity of directional information.The experimental results show that the detection accuracy of the proposed method on DOTA,HRSC2O16 and DIOR-R datasets reaches 76.88% , 98.43% and 65.06% ,respectively,which is 1.01% , 0.83% and 0.76% higher than that of the Oriented R-CNN method,respectively,which further verifies the efectiveness of background information and direction information collaboration.
Key Words: remote sensing images; rotating object detection; receptive field extending; orientationaware cross attention
1引言
遥感图像目标检测作为地理信息智能解译的核心任务,在灾害监测、城市规划及军事侦察等领域具有重要应用价值。(剩余21700字)