For a vision measurement system consisted of laser-CCD scanning sensors, an algorithm is proposed to extract and recognize the target object contour. Firstly, the two-dimensional（2D） point cloud that is output by th...For a vision measurement system consisted of laser-CCD scanning sensors, an algorithm is proposed to extract and recognize the target object contour. Firstly, the two-dimensional（2D） point cloud that is output by the integrated laser sensor is transformed into a binary image. Secondly, the potential target object contours are segmented and extracted based on the connected domain labeling and adaptive corner detection. Then, the target object contour is recognized by improved Hu invariant moments and BP neural network classifier. Finally, we extract the point data of the target object contour through the reverse transformation from a binary image to a 2D point cloud. The experimental results show that the average recognition rate is 98.5% and the average recognition time is 0.18 s per frame. This algorithm realizes the real-time tracking of the target object in the complex background and the condition of multi-moving objects.展开更多
The purpose of reverse engineering is to convert a large point cloud into a CAD model. In reverse engineering, the key issue is segmentation, i.e. studying how to subdivide the point cloud into smaller regions, where ...The purpose of reverse engineering is to convert a large point cloud into a CAD model. In reverse engineering, the key issue is segmentation, i.e. studying how to subdivide the point cloud into smaller regions, where each of them can be approximated by a single surface. Segmentation is relatively simple, if regions are bounded by sharp edges and small blends; problems arise when smoothly connected regions need to be separated. In this paper, a modified self-organizing feature map neural network （SOFM） is used to solve segmentation problem. Eight dimensional feature vectors （3-dimensional coordinates, 3-dimensional normal vectors, Gaussian curvature and mean curvature） are taken as input for SOFM. The weighted Euclidean distance measure is used to improve segmentation result. The method not only can deal with regions bounded by sharp edges, but also is very efficient to separating smoothly connected regions. The segmentation method using SOFM is robust to noise, and it operates directly on the point cloud. An examples is given to show the effect of SOFM algorithm.展开更多
The classification of point cloud data is the key technology of point cloud data information acquisition and 3D reconstruction, which has a wide range of applications. However, the existing point cloud classification ...The classification of point cloud data is the key technology of point cloud data information acquisition and 3D reconstruction, which has a wide range of applications. However, the existing point cloud classification methods have some shortcomings when extracting point cloud features, such as insufficient extraction of local information and overlooking the information in other neighborhood features in the point cloud, and not focusing on the point cloud channel information and spatial information. To solve the above problems, a point cloud classification network based on graph convolution and fusion attention mechanism is proposed to achieve more accurate classification results. Firstly, the point cloud is regarded as a node on the graph, the k-nearest neighbor algorithm is used to compose the graph and the information between points is dynamically captured by stacking multiple graph convolution layers;then, with the assistance of 2D experience of attention mechanism, an attention mechanism which has the capability to integrate more attention to point cloud spatial and channel information is introduced to increase the feature information of point cloud, aggregate local useful features and suppress useless features. Through the classification experiments on ModelNet40 dataset, the experimental results show that compared with PointNet network without considering the local feature information of the point cloud, the average classification accuracy of the proposed model has a 4.4% improvement and the overall classification accuracy has a 4.4% improvement. Compared with other networks, the classification accuracy of the proposed model has also been improved.展开更多
毫米波雷达凭借其出色的环境适应性、高分辨率和隐私保护等优势,在智能家居、智慧养老和安防监控等领域具有广泛的应用前景。毫米波雷达三维点云是一种重要的空间数据表达形式,对于人体行为姿态识别具有极大的价值。然而,由于毫米波雷...毫米波雷达凭借其出色的环境适应性、高分辨率和隐私保护等优势,在智能家居、智慧养老和安防监控等领域具有广泛的应用前景。毫米波雷达三维点云是一种重要的空间数据表达形式,对于人体行为姿态识别具有极大的价值。然而,由于毫米波雷达点云具有强稀疏性,给精准快速识别人体动作带来了巨大的挑战。针对这一问题,该文公开了一个毫米波雷达人体动作三维点云数据集mmWave-3DPCHM-1.0,并提出了相应的数据处理方法和人体动作识别模型。该数据集由TI公司的IWR1443-ISK和Vayyar公司的vBlu射频成像模组分别采集,包括常见的12种人体动作,如走路、挥手、站立和跌倒等。在网络模型方面,该文将边缘卷积(EdgeConv)与Transformer相结合,提出了一种处理长时序三维点云的网络模型,即Point EdgeConv and Transformer(PETer)网络。该网络通过边缘卷积对三维点云逐帧创建局部有向邻域图,以提取单帧点云的空间几何特征,并通过堆叠多个编码器的Transformer模块,提取多帧点云之间的时序关系。实验结果表明,所提出的PETer网络在所构建的TI数据集和Vayyar数据集上的平均识别准确率分别达到98.77%和99.51%,比传统最优的基线网络模型提高了大约5%,且网络规模仅为1.09 M,适于在存储受限的边缘设备上部署。展开更多
文摘For a vision measurement system consisted of laser-CCD scanning sensors, an algorithm is proposed to extract and recognize the target object contour. Firstly, the two-dimensional（2D） point cloud that is output by the integrated laser sensor is transformed into a binary image. Secondly, the potential target object contours are segmented and extracted based on the connected domain labeling and adaptive corner detection. Then, the target object contour is recognized by improved Hu invariant moments and BP neural network classifier. Finally, we extract the point data of the target object contour through the reverse transformation from a binary image to a 2D point cloud. The experimental results show that the average recognition rate is 98.5% and the average recognition time is 0.18 s per frame. This algorithm realizes the real-time tracking of the target object in the complex background and the condition of multi-moving objects.
基金Supported by the National Natural Science Foundation of China(60573177), the Aeronautical Science Foundation of China (04H53059) , the natural Science Foundation of Henan Province (200510078010) and Youth Science Foundation at North China Institute of Water Conservancy and Hydroelectric Power(HSQJ2004003)
文摘The purpose of reverse engineering is to convert a large point cloud into a CAD model. In reverse engineering, the key issue is segmentation, i.e. studying how to subdivide the point cloud into smaller regions, where each of them can be approximated by a single surface. Segmentation is relatively simple, if regions are bounded by sharp edges and small blends; problems arise when smoothly connected regions need to be separated. In this paper, a modified self-organizing feature map neural network （SOFM） is used to solve segmentation problem. Eight dimensional feature vectors （3-dimensional coordinates, 3-dimensional normal vectors, Gaussian curvature and mean curvature） are taken as input for SOFM. The weighted Euclidean distance measure is used to improve segmentation result. The method not only can deal with regions bounded by sharp edges, but also is very efficient to separating smoothly connected regions. The segmentation method using SOFM is robust to noise, and it operates directly on the point cloud. An examples is given to show the effect of SOFM algorithm.
文摘The classification of point cloud data is the key technology of point cloud data information acquisition and 3D reconstruction, which has a wide range of applications. However, the existing point cloud classification methods have some shortcomings when extracting point cloud features, such as insufficient extraction of local information and overlooking the information in other neighborhood features in the point cloud, and not focusing on the point cloud channel information and spatial information. To solve the above problems, a point cloud classification network based on graph convolution and fusion attention mechanism is proposed to achieve more accurate classification results. Firstly, the point cloud is regarded as a node on the graph, the k-nearest neighbor algorithm is used to compose the graph and the information between points is dynamically captured by stacking multiple graph convolution layers;then, with the assistance of 2D experience of attention mechanism, an attention mechanism which has the capability to integrate more attention to point cloud spatial and channel information is introduced to increase the feature information of point cloud, aggregate local useful features and suppress useless features. Through the classification experiments on ModelNet40 dataset, the experimental results show that compared with PointNet network without considering the local feature information of the point cloud, the average classification accuracy of the proposed model has a 4.4% improvement and the overall classification accuracy has a 4.4% improvement. Compared with other networks, the classification accuracy of the proposed model has also been improved.
文摘目的 基于点云的神经渲染方法受点云质量及特征提取的影响,易导致新视角合成图像渲染质量下降,为此提出一种融合局部空间信息的新视角合成方法。方法 针对点云质量及提取特征不足的问题,首先,设计一种神经点云特征对齐模块,将点云与图像匹配区域的特征进行对齐,融合后构成神经点云,提升其特征的局部表达能力;其次,提出一种神经点云Transformer模块,用于融合局部神经点云的上下文信息,在点云质量不佳的情况下仍能提取可靠的局部空间信息,有效增强了点云神经渲染方法的合成质量。结果 实验结果表明,在真实场景数据集中,对于只包含单一物品的数据集Tanks and Temples,本文方法在峰值信噪比(peak signal to noise ratio,PSNR)指标上与NeRF(neural radiance field)方法相比提升19.2%,相较于使用点云输入的方法 Tetra-NeRF和Point-NeRF分别提升了6.4%和3.8%,即使在场景更为复杂的ScanNet数据集中,与NeRF方法及Point-NeRF相比分别提升了34.6%和2.1%。结论 本文方法能够更好地利用点云的局部空间信息,有效改善了稀疏视角图像输入下因点云质量和提取特征导致的渲染质量下降,实验结果验证了本文方法的有效性。
文摘毫米波雷达凭借其出色的环境适应性、高分辨率和隐私保护等优势,在智能家居、智慧养老和安防监控等领域具有广泛的应用前景。毫米波雷达三维点云是一种重要的空间数据表达形式,对于人体行为姿态识别具有极大的价值。然而,由于毫米波雷达点云具有强稀疏性,给精准快速识别人体动作带来了巨大的挑战。针对这一问题,该文公开了一个毫米波雷达人体动作三维点云数据集mmWave-3DPCHM-1.0,并提出了相应的数据处理方法和人体动作识别模型。该数据集由TI公司的IWR1443-ISK和Vayyar公司的vBlu射频成像模组分别采集,包括常见的12种人体动作,如走路、挥手、站立和跌倒等。在网络模型方面,该文将边缘卷积(EdgeConv)与Transformer相结合,提出了一种处理长时序三维点云的网络模型,即Point EdgeConv and Transformer(PETer)网络。该网络通过边缘卷积对三维点云逐帧创建局部有向邻域图,以提取单帧点云的空间几何特征,并通过堆叠多个编码器的Transformer模块,提取多帧点云之间的时序关系。实验结果表明,所提出的PETer网络在所构建的TI数据集和Vayyar数据集上的平均识别准确率分别达到98.77%和99.51%,比传统最优的基线网络模型提高了大约5%,且网络规模仅为1.09 M,适于在存储受限的边缘设备上部署。
