Atmospheric pressure plasma-liquid interactions exist in a variety of applications,including wastewater treatment,wound sterilization,and disinfection.In practice,the phenomenon of liquid surface depression will inevi...Atmospheric pressure plasma-liquid interactions exist in a variety of applications,including wastewater treatment,wound sterilization,and disinfection.In practice,the phenomenon of liquid surface depression will inevitably appear.The applied gas will cause a depression on the liquid surface,which will undoubtedly affect the plasma generation and further affect the application performance.However,the effect of liquid surface deformation on the plasma is still unclear.In this work,numerical models are developed to reveal the mechanism of liquid surface depressions affecting plasma discharge characteristics and the consequential distribution of plasma species,and further study the influence of liquid surface depressions of different sizes generated by different helium flow rates on the plasma.Results show that the liquid surface deformation changes the initial spatial electric field,resulting in the rearrangement of electrons on the liquid surface.The charges deposited on the liquid surface further increase the degree of distortion of the electric field.Moreover,the electric field and electron distribution affected by the liquid surface depression significantly influence the generation and distribution of active species,which determines the practical effectiveness of the relevant applications.This work explores the phenomenon of liquid surface depression,which has been neglected in previous related work,and contributes to further understanding of plasma-liquid interactions,providing better theoretical guidance for related applications and technologies.展开更多
针对目前多视图三维重建无法适应高分辨率场景、重建完整性差、忽略全局背景信息等问题,提出一种融合可变形卷积与改进自注意力机制的三维重建网络MVFSAM-CasMVSNet。首先,设计专用于多视图立体重建任务的可变形卷积模块,自适应地调整...针对目前多视图三维重建无法适应高分辨率场景、重建完整性差、忽略全局背景信息等问题,提出一种融合可变形卷积与改进自注意力机制的三维重建网络MVFSAM-CasMVSNet。首先,设计专用于多视图立体重建任务的可变形卷积模块,自适应地调整提取特征的范围,增强深度突变的特征提取能力。其次,考虑到多视图间深度信息的关联性和特征交互,设计一种多视图融合自注意力模块,通过计算复杂度较低的线性自注意力聚合每个视图内部的远程上下文信息,并通过改进的多头注意力捕获参考视图与源视图间的深度依赖关系。最后利用多阶段策略构建匹配代价体并对其进行正则化,使用具有更高分辨率的代价体生成深度图。在DTU数据集上的测试结果显示,该网络与基准模型相比,完整性、准确性、整体性分别提升15.6%、7.4%、11.8%,与现有其他模型相比具有最优的整体性。同时,在Tanks and Temples数据集上的实验结果显示,该网络与基准模型相比平均F-score提升6.5%。该网络在多视图三维重建领域针对高分辨率场景具有优良的重建效果与泛化能力。展开更多
基金supported by National Natural Science Foundation of China(No.52377145).
文摘Atmospheric pressure plasma-liquid interactions exist in a variety of applications,including wastewater treatment,wound sterilization,and disinfection.In practice,the phenomenon of liquid surface depression will inevitably appear.The applied gas will cause a depression on the liquid surface,which will undoubtedly affect the plasma generation and further affect the application performance.However,the effect of liquid surface deformation on the plasma is still unclear.In this work,numerical models are developed to reveal the mechanism of liquid surface depressions affecting plasma discharge characteristics and the consequential distribution of plasma species,and further study the influence of liquid surface depressions of different sizes generated by different helium flow rates on the plasma.Results show that the liquid surface deformation changes the initial spatial electric field,resulting in the rearrangement of electrons on the liquid surface.The charges deposited on the liquid surface further increase the degree of distortion of the electric field.Moreover,the electric field and electron distribution affected by the liquid surface depression significantly influence the generation and distribution of active species,which determines the practical effectiveness of the relevant applications.This work explores the phenomenon of liquid surface depression,which has been neglected in previous related work,and contributes to further understanding of plasma-liquid interactions,providing better theoretical guidance for related applications and technologies.
文摘针对目前多视图三维重建无法适应高分辨率场景、重建完整性差、忽略全局背景信息等问题,提出一种融合可变形卷积与改进自注意力机制的三维重建网络MVFSAM-CasMVSNet。首先,设计专用于多视图立体重建任务的可变形卷积模块,自适应地调整提取特征的范围,增强深度突变的特征提取能力。其次,考虑到多视图间深度信息的关联性和特征交互,设计一种多视图融合自注意力模块,通过计算复杂度较低的线性自注意力聚合每个视图内部的远程上下文信息,并通过改进的多头注意力捕获参考视图与源视图间的深度依赖关系。最后利用多阶段策略构建匹配代价体并对其进行正则化,使用具有更高分辨率的代价体生成深度图。在DTU数据集上的测试结果显示,该网络与基准模型相比,完整性、准确性、整体性分别提升15.6%、7.4%、11.8%,与现有其他模型相比具有最优的整体性。同时,在Tanks and Temples数据集上的实验结果显示,该网络与基准模型相比平均F-score提升6.5%。该网络在多视图三维重建领域针对高分辨率场景具有优良的重建效果与泛化能力。
