检索规则说明:AND代表“并且”;OR代表“或者”;NOT代表“不包含”;(注意必须大写,运算符两边需空一格)
检 索 范 例 :范例一: (K=图书馆学 OR K=情报学) AND A=范并思 范例二:J=计算机应用与软件 AND (U=C++ OR U=Basic) NOT M=Visual
作 者:杨傲雷[1,2] 甘少英 杨帮华 苗中华[1] 徐昱琳[1] Yang Aolei;Gan Shaoying;Yang Banghua;Miao Zhonghua;Xu Yulin(School of Mechatronic Engineering and Automation,Shanghai University,Shanghai 200444,China;Shanghai Key Laboratory of Power Station Automation Technology,Shanghai 200444,China)
机构地区:[1]上海大学机电工程与自动化学院,上海200444 [2]上海市电站自动化技术重点实验室,上海200444
出 处:《仪器仪表学报》2024年第12期179-189,共11页Chinese Journal of Scientific Instrument
基 金:国家重点研发计划(2023YFF1203503);上海市自然科学基金(22ZR1424200)项目资助。
摘 要:针对移动机器人相机位姿估计问题,提出一种基于混合频域Transformer的相机位姿估计方法,旨在从RGB图像中预测相机的位置与方向。首先,构建了室内场景数据集RotIndoor,每个样本包含场景RGB图像和通过VICON系统获取的相机位姿真值;其次,提出位姿回归网络模型CamPose,该模型融合空间域和频域的信息,提升了图像特征表达能力,进而实现高精度的相机位姿估计。具体而言,CamPose引入基于差分卷积网络的特征增强模块,捕获图像细粒度特征;设计了频域编码层,通过傅里叶变换提取频率特征,并整合频域注意力模块,使模型感知不同频率成分的重要性。最后,在公开数据集7Scenes和RotIndoor上进行了实验验证表明,该方法在7Scenes数据集上的位姿估计误差为0.17 m/7.85°,在RotIndoor上定位精度提高了23%。To address the challenges of camera pose estimation and mobile robot localization,a camera pose estimation method is proposed based on a hybrid frequency domain Transformer to predict the position and orientation of a camera from RGB images.Firstly,a camera pose estimation dataset,RotIndoor,is constructed based on indoor scenes,with each sample containing an RGB image of the scene and the ground truth camera poses obtained from a VICON system.Secondly,a pose regression network model,CamPose,is introduced,which effectively integrates spatial and frequency domain information to enhance the representation capability of image features,ultimately achieving higher accuracy in camera pose estimation.Specifically,CamPose incorporates a feature enhancement module based on differential convolution networks to capture fine-grained features within the images.Additionally,a frequency domain encoding layer is designed that applies Fourier transformation to extract frequency characteristics while integrating a frequency domain attention module,enabling the model to sensitively perceive the importance of different frequency components.Finally,experiments are implemented on the public datasets 7Scenes and RotIndoor.The experimental results show that the pose estimation error on the 7Scenes dataset is reduced to 0.17 m/7.85°,and the positioning accuracy on RotIndoor is improved by 23%compared to other methods.Addressing the challenges of camera pose estimation and mobile robot localization,proposes a camera pose estimation method based on a hybrid frequency domain Transformer,aimed at predicting the position and orientation of a camera from RGB images.Firstly,a camera pose estimation dataset,RotIndoor,was constructed based on indoor scenes,with each sample containing an RGB image of the scene and the ground truth camera poses obtained from a VICON system.Secondly,a pose regression network model,CamPose,is introduced,which effectively integrates spatial and frequency domain information to enhance the representation capability of i
关 键 词:相机位姿估计 深度学习 特征增强 频域编码 TRANSFORMER
分 类 号:TP391[自动化与计算机技术—计算机应用技术] TH86[自动化与计算机技术—计算机科学与技术]
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在链接到云南高校图书馆文献保障联盟下载...
云南高校图书馆联盟文献共享服务平台 版权所有©
您的IP:216.73.216.222