针对现有基于伪点云的3D目标检测算法精度远低于基于真实激光雷达(Light Detection and ranging,LiDar)点云的3D目标检测,本文研究伪点云重构,并提出适合伪点云的3D目标检测网络.考虑到由图像深度转换得到的伪点云稠密且随深度增大逐渐...针对现有基于伪点云的3D目标检测算法精度远低于基于真实激光雷达(Light Detection and ranging,LiDar)点云的3D目标检测,本文研究伪点云重构,并提出适合伪点云的3D目标检测网络.考虑到由图像深度转换得到的伪点云稠密且随深度增大逐渐稀疏,本文提出深度相关伪点云稀疏化方法,在减少后续计算量的同时保留中远距离更多的有效伪点云,实现伪点云重构.本文提出LiDar点云指导下特征分布趋同与语义关联的3D目标检测网络,在网络训练时引入LiDar点云分支来指导伪点云目标特征的生成,使生成的伪点云特征分布趋同于LiDar点云特征分布,从而降低数据源不一致造成的检测性能损失;针对RPN(Region Proposal Network)网络获取的3D候选框内的伪点云间语义关联不足的问题,设计注意力感知模块,在伪点云特征表示中通过注意力机制嵌入点间的语义关联关系,提升3D目标检测精度.在KITTI 3D目标检测数据集上的实验结果表明:现有的3D目标检测网络采用重构后的伪点云,检测精度提升了2.61%;提出的特征分布趋同与语义关联的3D目标检测网络,将基于伪点云的3D目标检测精度再提升0.57%,相比其他优秀的3D目标检测方法在检测精度上也有提升.展开更多
Recognition of human gesture actions is a challenging issue due to the complex patterns in both visual andskeletal features. Existing gesture action recognition (GAR) methods typically analyze visual and skeletal data...Recognition of human gesture actions is a challenging issue due to the complex patterns in both visual andskeletal features. Existing gesture action recognition (GAR) methods typically analyze visual and skeletal data,failing to meet the demands of various scenarios. Furthermore, multi-modal approaches lack the versatility toefficiently process both uniformand disparate input patterns.Thus, in this paper, an attention-enhanced pseudo-3Dresidual model is proposed to address the GAR problem, called HgaNets. This model comprises two independentcomponents designed formodeling visual RGB (red, green and blue) images and 3Dskeletal heatmaps, respectively.More specifically, each component consists of two main parts: 1) a multi-dimensional attention module forcapturing important spatial, temporal and feature information in human gestures;2) a spatiotemporal convolutionmodule that utilizes pseudo-3D residual convolution to characterize spatiotemporal features of gestures. Then,the output weights of the two components are fused to generate the recognition results. Finally, we conductedexperiments on four datasets to assess the efficiency of the proposed model. The results show that the accuracy onfour datasets reaches 85.40%, 91.91%, 94.70%, and 95.30%, respectively, as well as the inference time is 0.54 s andthe parameters is 2.74M. These findings highlight that the proposed model outperforms other existing approachesin terms of recognition accuracy.展开更多
The developing processes of stress and deformation fields of a protected layer after mining an upper-protective layer with a bow pseudo-incline technique were simulated to locate the protection region. The pressure re...The developing processes of stress and deformation fields of a protected layer after mining an upper-protective layer with a bow pseudo-incline technique were simulated to locate the protection region. The pressure relief of the protected layer was analyzed after mining the upper-protective layer. The pressure relief angle along the strike and incline were located according to the rules of protection of the deformation and stress pressure-relief of the protective layer after mining. This results show that the upper-protective layer with the bow pseudo-incline technique have an upper and downside pressure relief angle of 85 and 68 degrees respectively; the distribution of strike pressure relief angles along the pseudo-incline working face is uneven and their values range from 38.3 to 51 degrees. The pressure relief angle of the inclined middle location was the largest. The distribution of the protection region of the upper-protective layer with the bow pseudo-incline technique located by practical tests and numerical simulation is essentially consistent, compared with the results obtained by these methods.展开更多
With 3D orthogonal and pseudo-orthogonal weaves, woven sructures with lengthwise and widthwise changing cross section on one side or both sides of the structure can be constructed. The weave formation and the looming ...With 3D orthogonal and pseudo-orthogonal weaves, woven sructures with lengthwise and widthwise changing cross section on one side or both sides of the structure can be constructed. The weave formation and the looming draft creation are discussed in this paper which can be used as references to manufacture woven preforms with changing cross sections.展开更多
基金the National Natural Science Foundation of China under Grant No.62072255.
文摘Recognition of human gesture actions is a challenging issue due to the complex patterns in both visual andskeletal features. Existing gesture action recognition (GAR) methods typically analyze visual and skeletal data,failing to meet the demands of various scenarios. Furthermore, multi-modal approaches lack the versatility toefficiently process both uniformand disparate input patterns.Thus, in this paper, an attention-enhanced pseudo-3Dresidual model is proposed to address the GAR problem, called HgaNets. This model comprises two independentcomponents designed formodeling visual RGB (red, green and blue) images and 3Dskeletal heatmaps, respectively.More specifically, each component consists of two main parts: 1) a multi-dimensional attention module forcapturing important spatial, temporal and feature information in human gestures;2) a spatiotemporal convolutionmodule that utilizes pseudo-3D residual convolution to characterize spatiotemporal features of gestures. Then,the output weights of the two components are fused to generate the recognition results. Finally, we conductedexperiments on four datasets to assess the efficiency of the proposed model. The results show that the accuracy onfour datasets reaches 85.40%, 91.91%, 94.70%, and 95.30%, respectively, as well as the inference time is 0.54 s andthe parameters is 2.74M. These findings highlight that the proposed model outperforms other existing approachesin terms of recognition accuracy.
基金Projects PLN0610 supported by the Open Fund of State Key Lab of Oil and Gas Reservoir Geology and Exploitation (Southwest Petroleum University)HKLGF200706 by the Opening Project of Henan Key Laboratory of Coal Mine Methane and Fire Prevention+3 种基金50334060, 50474025 and 50774106 by the National Natural Science Foundation of China2005CB221502 by the National Basic Research Program of China50621403 by the Natural Science Innova-tion Group Foundation of ChinaCSTC, 2006BB7147, 2006AA7002 by the Natural Science Foundation of Chongqing
文摘The developing processes of stress and deformation fields of a protected layer after mining an upper-protective layer with a bow pseudo-incline technique were simulated to locate the protection region. The pressure relief of the protected layer was analyzed after mining the upper-protective layer. The pressure relief angle along the strike and incline were located according to the rules of protection of the deformation and stress pressure-relief of the protective layer after mining. This results show that the upper-protective layer with the bow pseudo-incline technique have an upper and downside pressure relief angle of 85 and 68 degrees respectively; the distribution of strike pressure relief angles along the pseudo-incline working face is uneven and their values range from 38.3 to 51 degrees. The pressure relief angle of the inclined middle location was the largest. The distribution of the protection region of the upper-protective layer with the bow pseudo-incline technique located by practical tests and numerical simulation is essentially consistent, compared with the results obtained by these methods.
文摘With 3D orthogonal and pseudo-orthogonal weaves, woven sructures with lengthwise and widthwise changing cross section on one side or both sides of the structure can be constructed. The weave formation and the looming draft creation are discussed in this paper which can be used as references to manufacture woven preforms with changing cross sections.
