Self-supervised monocular depth estimation has emerged as a major research focus in recent years,primarily due to the elimination of ground-truth depth dependence.However,the prevailing architectures in this domain su...Self-supervised monocular depth estimation has emerged as a major research focus in recent years,primarily due to the elimination of ground-truth depth dependence.However,the prevailing architectures in this domain suffer from inherent limitations:existing pose network branches infer camera ego-motion exclusively under static-scene and Lambertian-surface assumptions.These assumptions are often violated in real-world scenarios due to dynamic objects,non-Lambertian reflectance,and unstructured background elements,leading to pervasive artifacts such as depth discontinuities(“holes”),structural collapse,and ambiguous reconstruction.To address these challenges,we propose a novel framework that integrates scene dynamic pose estimation into the conventional self-supervised depth network,enhancing its ability to model complex scene dynamics.Our contributions are threefold:(1)a pixel-wise dynamic pose estimation module that jointly resolves the pose transformations of moving objects and localized scene perturbations;(2)a physically-informed loss function that couples dynamic pose and depth predictions,designed to mitigate depth errors arising from high-speed distant objects and geometrically inconsistent motion profiles;(3)an efficient SE(3)transformation parameterization that streamlines network complexity and temporal pre-processing.Extensive experiments on the KITTI and NYU-V2 benchmarks show that our framework achieves state-of-the-art performance in both quantitative metrics and qualitative visual fidelity,significantly improving the robustness and generalization of monocular depth estimation under dynamic conditions.展开更多
With the rapid development of space science projects,large deployable mechanisms have been widely used.However,due to the effects of mechanical friction and gravitational acceleration,on-orbit mechanisms cannot be alw...With the rapid development of space science projects,large deployable mechanisms have been widely used.However,due to the effects of mechanical friction and gravitational acceleration,on-orbit mechanisms cannot be always deployed to the expected pose.For some precision optical mechanisms,even a minor deviation can result in significant error,so it needs to be measured and corrected.In this paper,the deployment process was modeled and simplified as rotation under single-rotation-axis constraint and translation under single-direction constraint.To solve the problem,a method based on cross-ratio invariability was proposed.The proposed method does not rely on camera calibration techniques,as well as artificial marking points,both of which are necessary in PnP.Instead,only three calibration images before launch and a measurement image on orbit were required.Simulations and experiments demonstrated that the proposed method is more accurate than PnP.In addition,experiments also proved that the feasibility of the proposed method under dark conditions with the aid of a light source and some reflective marking points.展开更多
We present iterative numerical methods for solving the inverse problem of recovering the nonnegative Robin coefficient from partial boundary measurement of the solution to the Laplace equation. Based on the boundary i...We present iterative numerical methods for solving the inverse problem of recovering the nonnegative Robin coefficient from partial boundary measurement of the solution to the Laplace equation. Based on the boundary integral equation formulation of the problem, nonnegativity constraints in the form of a penalty term are incorporated conveniently into least-squares iteration schemes for solving the inverse problem. Numerical implementation and examples are presented to illustrate the effectiveness of this strategy in improving recovery results.展开更多
目的 针对目前三维人体姿态估计方法未能有效处理时间序列冗余,难以捕获人体关节上微小变化的问题,提出一种融合多关节特征的单目视觉三维人体姿态估计网络。方法 在关节运动特征提取模块中,采用多分支操作提取关节在时间维度上的运动特...目的 针对目前三维人体姿态估计方法未能有效处理时间序列冗余,难以捕获人体关节上微小变化的问题,提出一种融合多关节特征的单目视觉三维人体姿态估计网络。方法 在关节运动特征提取模块中,采用多分支操作提取关节在时间维度上的运动特征,并将不同特征融合形成具有高度表达力的特征表示。关节特征融合模块整合了不同关节组和中间帧的全局信息,通过矩阵内积的方式表达不同关节组在高纬度空间的相对位置及相互联系,得到中间3D姿态的初估值。关节约束模块引入中间帧的2D关节点空间位置关系作为隐式约束,与中间帧3D姿态初估值融合,减少不合理的姿态输出,提高最终3D姿态估计的准确性。结果 实验结果表明,与MHFormer方法相比,本文方法在Human3.6M数据集上的平均关节位置误差(mean per joint position error,MPJPE)结果为29.0 mm,误差降低4.9%,对于复杂动作,如SittingDown和WalkDog,误差降低了7.7%和8.2%。在MPI-INF-3DHP数据集上,MPJPE指标降低36.2%,曲线下面积(area under the curve,AUC)指标提升12.9%,正确关节点百分比(percentage of correct keypoints,PCK)指标提升3%。实验结果体现出在面对复杂动作问题时,网络利用各分支提取了不同的关节时序运动特征,将不同关节组的位置信息进行融合交互,结合当前帧的关节姿态信息加以约束,取得更高的精度。在HumanEva数据集上的实验结果表明了本文方法适用不同数据集,消融实验进一步验证了各个模块的有效性。结论 本文网络有效地融合了人体多关节特征,可以更好地提高单目视觉三维人体姿态估计的准确性,且具备较高的泛化性。展开更多
目的现有表情识别方法聚焦提升模型的整体识别准确率,对方法的头部姿态鲁棒性研究不充分。在实际应用中,人的头部姿态往往变化多样,影响表情识别效果,因此研究头部姿态对表情识别的影响,并提升模型在该方面的鲁棒性显得尤为重要。为此,...目的现有表情识别方法聚焦提升模型的整体识别准确率,对方法的头部姿态鲁棒性研究不充分。在实际应用中,人的头部姿态往往变化多样,影响表情识别效果,因此研究头部姿态对表情识别的影响,并提升模型在该方面的鲁棒性显得尤为重要。为此,在深入分析头部姿态对表情识别影响的基础上,提出一种能够基于无标签非正脸表情数据提升模型头部姿态鲁棒性的半监督表情识别方法。方法首先按头部姿态对典型表情识别数据集AffectNet重新划分,构建了AffectNet-Yaw数据集,支持在不同角度上进行模型精度测试,提升了模型对比公平性。其次,提出一种基于双一致性约束的半监督表情识别方法(dual-consistency semi-supervised learning for facial expression recognition,DCSSL),利用空间一致性模块对翻转前后人脸图像的类别激活一致性进行空间约束,使模型训练时更关注面部表情关键区域特征;利用语义一致性模块通过非对称数据增强和自学式学习方法不断地筛选高质量非正脸数据用于模型优化。在无需对非正脸表情数据人工标注的情况下,方法直接从有标签正脸数据和无标签非正脸数据中学习。最后,联合优化了交叉熵损失、空间一致性约束损失和语义一致性约束损失函数,以确保有监督学习和半监督学习之间的平衡。结果实验结果表明,头部姿态对自然场景表情识别有显著影响;提出AffectNet-Yaw具有更均衡的头部姿态分布,有效促进了对这种影响的全面评估;DCSSL方法结合空间一致性和语义一致性约束充分利用无标签非正脸表情数据,显著提高了模型在头部姿态变化下的鲁棒性,较MA-NET(multi-scale and local attention network)和EfficientFace全监督方法,平均表情识别精度分别提升了5.40%和17.01%。结论本文提出的双一致性半监督方法能充分利用正脸和非正脸数据,显著提升了模型在头部姿态变化下的表情识别精度;新数据集有效支撑了对头部姿态对表情识别影响的全面评估。展开更多
基金supported in part by the National Natural Science Foundation of China under Grants 62071345。
文摘Self-supervised monocular depth estimation has emerged as a major research focus in recent years,primarily due to the elimination of ground-truth depth dependence.However,the prevailing architectures in this domain suffer from inherent limitations:existing pose network branches infer camera ego-motion exclusively under static-scene and Lambertian-surface assumptions.These assumptions are often violated in real-world scenarios due to dynamic objects,non-Lambertian reflectance,and unstructured background elements,leading to pervasive artifacts such as depth discontinuities(“holes”),structural collapse,and ambiguous reconstruction.To address these challenges,we propose a novel framework that integrates scene dynamic pose estimation into the conventional self-supervised depth network,enhancing its ability to model complex scene dynamics.Our contributions are threefold:(1)a pixel-wise dynamic pose estimation module that jointly resolves the pose transformations of moving objects and localized scene perturbations;(2)a physically-informed loss function that couples dynamic pose and depth predictions,designed to mitigate depth errors arising from high-speed distant objects and geometrically inconsistent motion profiles;(3)an efficient SE(3)transformation parameterization that streamlines network complexity and temporal pre-processing.Extensive experiments on the KITTI and NYU-V2 benchmarks show that our framework achieves state-of-the-art performance in both quantitative metrics and qualitative visual fidelity,significantly improving the robustness and generalization of monocular depth estimation under dynamic conditions.
基金Supported in part by the National Natural Science Foundation of China(No.62271148).
文摘With the rapid development of space science projects,large deployable mechanisms have been widely used.However,due to the effects of mechanical friction and gravitational acceleration,on-orbit mechanisms cannot be always deployed to the expected pose.For some precision optical mechanisms,even a minor deviation can result in significant error,so it needs to be measured and corrected.In this paper,the deployment process was modeled and simplified as rotation under single-rotation-axis constraint and translation under single-direction constraint.To solve the problem,a method based on cross-ratio invariability was proposed.The proposed method does not rely on camera calibration techniques,as well as artificial marking points,both of which are necessary in PnP.Instead,only three calibration images before launch and a measurement image on orbit were required.Simulations and experiments demonstrated that the proposed method is more accurate than PnP.In addition,experiments also proved that the feasibility of the proposed method under dark conditions with the aid of a light source and some reflective marking points.
文摘We present iterative numerical methods for solving the inverse problem of recovering the nonnegative Robin coefficient from partial boundary measurement of the solution to the Laplace equation. Based on the boundary integral equation formulation of the problem, nonnegativity constraints in the form of a penalty term are incorporated conveniently into least-squares iteration schemes for solving the inverse problem. Numerical implementation and examples are presented to illustrate the effectiveness of this strategy in improving recovery results.
文摘目的 针对目前三维人体姿态估计方法未能有效处理时间序列冗余,难以捕获人体关节上微小变化的问题,提出一种融合多关节特征的单目视觉三维人体姿态估计网络。方法 在关节运动特征提取模块中,采用多分支操作提取关节在时间维度上的运动特征,并将不同特征融合形成具有高度表达力的特征表示。关节特征融合模块整合了不同关节组和中间帧的全局信息,通过矩阵内积的方式表达不同关节组在高纬度空间的相对位置及相互联系,得到中间3D姿态的初估值。关节约束模块引入中间帧的2D关节点空间位置关系作为隐式约束,与中间帧3D姿态初估值融合,减少不合理的姿态输出,提高最终3D姿态估计的准确性。结果 实验结果表明,与MHFormer方法相比,本文方法在Human3.6M数据集上的平均关节位置误差(mean per joint position error,MPJPE)结果为29.0 mm,误差降低4.9%,对于复杂动作,如SittingDown和WalkDog,误差降低了7.7%和8.2%。在MPI-INF-3DHP数据集上,MPJPE指标降低36.2%,曲线下面积(area under the curve,AUC)指标提升12.9%,正确关节点百分比(percentage of correct keypoints,PCK)指标提升3%。实验结果体现出在面对复杂动作问题时,网络利用各分支提取了不同的关节时序运动特征,将不同关节组的位置信息进行融合交互,结合当前帧的关节姿态信息加以约束,取得更高的精度。在HumanEva数据集上的实验结果表明了本文方法适用不同数据集,消融实验进一步验证了各个模块的有效性。结论 本文网络有效地融合了人体多关节特征,可以更好地提高单目视觉三维人体姿态估计的准确性,且具备较高的泛化性。
文摘为提高6轴机器人的运动学绝对定位精度,提出了一种闭环校准方法.首先,通过视觉位姿测量方法,建立多点位姿约束;其次,利用差分方法构建辨识方程;最后,采用加权最小二乘法对机器人运动学参数进行辨识.测量设备为安装在机器人末端的相机,其与固定的高精度标定板配合使用,以实现多点位姿约束.根据工业应用中的逆解参数约束条件和手眼系统的运动学耦合关系,确定了包含14个参数的最小校准参数集.基于此参数集,应用加权最小二乘法进行参数辨识,从而完成了机器人的运动学参数校准.采用该方法对华数HSR-JR605机器人进行校准后,机器人末端绕点运动的平均距离误差显著降低,从2.77 mm降低到0.42 mm.
文摘目的现有表情识别方法聚焦提升模型的整体识别准确率,对方法的头部姿态鲁棒性研究不充分。在实际应用中,人的头部姿态往往变化多样,影响表情识别效果,因此研究头部姿态对表情识别的影响,并提升模型在该方面的鲁棒性显得尤为重要。为此,在深入分析头部姿态对表情识别影响的基础上,提出一种能够基于无标签非正脸表情数据提升模型头部姿态鲁棒性的半监督表情识别方法。方法首先按头部姿态对典型表情识别数据集AffectNet重新划分,构建了AffectNet-Yaw数据集,支持在不同角度上进行模型精度测试,提升了模型对比公平性。其次,提出一种基于双一致性约束的半监督表情识别方法(dual-consistency semi-supervised learning for facial expression recognition,DCSSL),利用空间一致性模块对翻转前后人脸图像的类别激活一致性进行空间约束,使模型训练时更关注面部表情关键区域特征;利用语义一致性模块通过非对称数据增强和自学式学习方法不断地筛选高质量非正脸数据用于模型优化。在无需对非正脸表情数据人工标注的情况下,方法直接从有标签正脸数据和无标签非正脸数据中学习。最后,联合优化了交叉熵损失、空间一致性约束损失和语义一致性约束损失函数,以确保有监督学习和半监督学习之间的平衡。结果实验结果表明,头部姿态对自然场景表情识别有显著影响;提出AffectNet-Yaw具有更均衡的头部姿态分布,有效促进了对这种影响的全面评估;DCSSL方法结合空间一致性和语义一致性约束充分利用无标签非正脸表情数据,显著提高了模型在头部姿态变化下的鲁棒性,较MA-NET(multi-scale and local attention network)和EfficientFace全监督方法,平均表情识别精度分别提升了5.40%和17.01%。结论本文提出的双一致性半监督方法能充分利用正脸和非正脸数据,显著提升了模型在头部姿态变化下的表情识别精度;新数据集有效支撑了对头部姿态对表情识别影响的全面评估。
