基于Point Transformer方法的鱼类三维点云模型分类

为实现对不同鱼类的精准分类,研究共采集110尾真实鱼类的三维模型,对获取的3D模型进行基于预处理、旋转增强和下采样等操作后,获取了1650尾实验样本。然后基于Point Transformer网络和2个三维分类的对比网络进行数据集的分类训练和验证。结果表明,利用本实验的目标方法Point Transformer获得了比2个对比网络更好的分类结果,整体的分类准确率能够达到91.9%。同时对所使用的三维分类网络进行有效性评估,3个模型对于5种真实鱼类模型的分类是有意义的,其中Point Transformer的模型ROC曲线准确率最高,AUC面积最大,对于三维鱼类数据集的分类最为有效。研究提供了一种可以实现对鱼类三维模型进行精准分类的方法,为以后的智能化渔业资源监测提供一种新的技术手段。

Optimizing electronic structure through point defect engineering for enhanced electrocatalytic energy conversion

Point defect engineering endows catalysts with novel physical and chemical properties,elevating their electrocatalytic efficiency.The introduction of defects emerges as a promising strategy,effectively modifying the electronic structure of active sites.This optimization influences the adsorption energy of intermediates,thereby mitigating reaction energy barriers,altering paths,enhancing selectivity,and ultimately improving the catalytic efficiency of electrocatalysts.To elucidate the impact of defects on the electrocatalytic process,we comprehensively outline the roles of various point defects,their synthetic methodologies,and characterization techniques.Importantly,we consolidate insights into the relationship between point defects and catalytic activity for hydrogen/oxygen evolution and CO_(2)/O_(2)/N_(2) reduction reactions by integrating mechanisms from diverse reactions.This underscores the pivotal role of point defects in enhancing catalytic performance.At last,the principal challenges and prospects associated with point defects in current electrocatalysts are proposed,emphasizing their role in advancing the efficiency of electrochemical energy storage and conversion materials.

基于注意力权重PointNet++的电力走廊点云语义分割研究

传统的电力走廊点云数据的分割会出现精度低、数据局部特征捕获存在局限性等问题,为此提出了一种基于注意力权重的PointNet++网络场景分割模型。将深度学习中的PointNet++算法用于电力走廊场景分割中,再引入了空间注意力机制,帮助模型更有效地关注重要的空间区域。为此采用自制的数据集,并基于PointNet++网络模型的经典结构,在每个点集抽取(set abstraction,SA)模块中的多层感知机(multi layer perceptron,MLP)加入倒置瓶颈设计,提高对点云数据的处理效率和准确性。研究结果表明,与传统的PointNet++网络相比,改进的PointNet++网络平均交并比(mean intersection over union,mIoU)高出6.3%,加入空间注意力机制的改进模型在自制数据集上表现出更好的分割效果,尤其是在边界划分方面提升明显,验证了该方法在点云语义分割上的有效性。

基于PointNet的钢板毛坯垛点云分割

钢板毛坯垛拆垛工序中,钢板毛坯分层厚度的估计是推钢机准确且安全执行推钢动作的关键;当前不少企业此道生产工序仍主要靠操作员人工观察的方式估算钢板厚度及与传送辊道的相对高度,其估算不准易导致碰撞,造成输送设备损坏生产中断。提出一种钢板毛坯垛智能分层方法,该方法结合现场工况环境采用激光雷达,对钢板毛坯垛进行三维点云成像,然后对采集的点云数据,用PointNet神经网络框架进行特征识别、分层分割与提取,最后对分割的不同层根据标定值,换算成真实厚度。根据现场实验结果表明,PointNet对钢板毛坯垛分割的识别率达到了87.4%,厚度估算误差低于1.2cm,根据钢厂钢板毛坯规格表(3种规格150、160、180cm)可以准确估计出钢板毛坯厚度的规格,识别速度15帧/s,满足现场工况要求。

基于轻量化PointNet网络的林果园喷雾作业靶标实时识别方法

为了进一步提高喷雾机器人靶标检测的精准性、实时性和应用部署的实用性,该研究提出一种基于轻量化PointNet网络的林果园喷雾作业靶标实时识别方法。首先通过区域提取降采样、地面分割和改进DBSCAN聚类等点云预处理方法提取原始点云中的靶标;然后通过移动最小二乘上采样将靶标点云转化为满足点云识别网络输入要求的点云数据;最终通过在PointNet网络中引入残差模块和改进循环剪枝算法轻量化PointNet网络,完成林果树靶标的实时识别。试验结果表明,在ModelNet40数据集上,轻量化PointNet网络可达89.7%的准确率;在实际苗圃环境的试验中,该研究方法对靶标的识别准确率可达92.49%,同时误识率与拒识率分别为13.4%和6.47%,相较PointNet网络识别准确率提升了4.38个百分点,误识率和拒识率分别降低了7.2和4.07个百分点;轻量化PointNet网络识别准确率仅比PointNet++网络低1.14个百分点,误识率和拒识率分别高了0.9和1.12个百分点。但是轻量化PointNet网络的模型参数量较PointNet网络和PointNet++网络的模型参数量显著减少,仅为PointNet网络的11.5%,PointNet++网络的27.02%;运算量相较PointNet网络、PointNet++网络分别减少13.3和76.79个百分点。该研究提出的轻量化PointNet网络具有较高的实时性、精确性和鲁棒性,能够满足林果园喷雾作业的靶标识别需求,可为林果园喷雾作业靶标实时识别提供参考。

基于K F-PointNet++的油菜植株点云分割算法

针对传统的点云分割算法精度低、鲁棒性差的问题,提出了基于K近邻算法和特征融合的深度点云网络(point clouds network++based on K-nearest neighbor algorithm and feature fusion,K F-PointNet++)三维点云分割算法。该算法首先采用了K近邻(K-nearest neighbors,K NN)算法对点云进行分组;其次,将点云网络(point clouds network,PointNet)中的局部特征与中心点全局特征进行拼接,增强算法对几何细节和全局上下文信息的捕捉能力,从而提高算法的分割精度和鲁棒性,实现了对油菜点云器官的精准分割。使用自制的油菜点云数据集进行实验,结果表明,K F-PointNet++算法在油菜点云分割中的总体精度(overall accuracy,OA)可达97.1%,平均交并比(mean intersection over union,mIoU)为86.4%。该算法在分割性能方面明显优于PointNet、深度点云网络(PointNet++)和核点卷积(kernal point convolution,KPConv),可以为油菜表型研究提供可靠基础。

基于PointNet++的邻域特征增强点云语义分割方法

随着智能驾驶、机器人导航等以点云为基础的应用蓬勃发展,点云语义分割逐渐成为研究热点。然而,现有的点云语义分割方法存在局部特征提取不充分、特征融合不完整的缺陷。针对这些不足,提出了对应的解决方案。对于局部特征提取不充分的现象,通过嵌入邻域点的坐标、方向、距离等相关信息去关联邻域点的显式特征。对于特征融合不完整的现象,提出了一种最大池化与自注意力池化相结合的混合池化方法。网络架构基于PointNet++,并结合提出的局部特征提取和融合方法,在S3DIS数据集上的实验结果表明,与基线方法PointNet++相比,各评价指标都有不同程度的提高,证实了新方法的有效性和优越性。

基于Point Transformer v2的点云枝叶分离方法研究

准确高效的点云枝叶分离对精确计算森林树木的垂直参数至关重要。然而,当前的研究方法计算成本高,且依赖先验知识导致泛化能力不足。针对以上问题,文章提出利用基于点特征的Transformer网络进行自动化的森林场景三维点云的枝叶分离研究。该方法使用Point Transformer v2网络,首先利用网格编码模块提取可学习的局部结构关系,保留点云的几何拓扑结构;其次使用分组注意力实现多通道联合学习,降低特征的冗余度,提高计算的效率;最后构建了基于点的Transformer网络实现高精度森林树木三维点云语义分割,降低了对于先验知识的需求。使用地基激光扫描仪获取的加拿大和芬兰7个不同树种样地的三维点云数据,进行枝叶分离实验和精度评价。实验结果表明,网络整体精度(OA)为94.42%,mIoU为78.89%,能够适应不同树种、不同点云密度的森林场景的枝叶分离。

基于改进PointNet++的船体分段合拢面构件智能识别算法研究

[目的]三维扫描仪获得的船体分段合拢面点云数据,具有精度高、数据量大的优势,能够很好地反映分段合拢面的建造状况。由于现有的PointNet++网络无法处理大容量点云数据,因此提出一种基于改进PointNet++的船体分段合拢面构件智能识别算法,实现针对大容量船体分段合拢面点云数据构件的智能识别。[方法]基于超体素生长理论对船体分段合拢面点云数据进行分割及简化,构建船体分段合拢面点云数据集,并使用该数据集训练基于深度学习理论改进的PointNet++网络。[结果]网络模型在船体分段合拢面点云数据训练集和测试集上的收敛结果趋于稳定,在测试集上识别准确率达到90.012%。[结论]该方法具有良好的识别能力,能够完成船体分段合拢面构件的智能识别。

基于PointNet++进行附属设施语义分割的隧道收敛变形分析

随着城市轨道交通日趋广泛,隧道结构变形引起的地铁安全事故凸显,亟需对运营期隧道进行变形检测。隧道衬砌作为隧道变形分析的研究对象,衬砌内表面存在的大量附属设施影响隧道收敛变形分析精度。为了提高变形分析精度,解决点云处理环节中存在的自动化程度低的问题,提出基于PointNet++点云语义分割的隧道收敛变形分析方法。首先利用深度学习方法进行点云语义分割,对隧道衬砌附属设施进行自动滤除。然后对隧道衬砌进行断面提取,利用随机抽样一致性算法(Random Sample Consensus,RANSAC)对隧道断面点云进行采样,分析隧道收敛变形程度,从Z+F PROFILER 9012A激光断面扫描仪获取山东省济南市地铁盾构隧道点云实测数据上并进行应用。研究结果表明:所提出的处理方法可以有效地将大规模隧道衬砌与连接紧密的附属设施分离出来,隧道附属设施总体分类精度达到96%,滤波结果较好地保留了隧道衬砌原始形态特征。在对隧道整体和局部收敛变形分析的重复性验证中,测试区间内隧道整体变形精度往返测长半轴平均偏差为1.04 mm,短半轴平均偏差为0.9 mm,测试区间内隧道局部收敛变形往返测标准差最小为0.773 mm,最大为0.938 mm,可以满足隧道收敛变形分析的精度需求。研究结果可以有效提升处理大规模隧道数据的自动化程度,具有良好的有效性与可靠性,对运营期地铁隧道收敛变形检测或监测有较好的实践应用意义。

Point Cloud Classification Using Content-Based Transformer via Clustering in Feature Space

Recently, there have been some attempts of Transformer in 3D point cloud classification. In order to reduce computations, most existing methods focus on local spatial attention,but ignore their content and fail to establish relationships between distant but relevant points. To overcome the limitation of local spatial attention, we propose a point content-based Transformer architecture, called PointConT for short. It exploits the locality of points in the feature space(content-based), which clusters the sampled points with similar features into the same class and computes the self-attention within each class, thus enabling an effective trade-off between capturing long-range dependencies and computational complexity. We further introduce an inception feature aggregator for point cloud classification, which uses parallel structures to aggregate high-frequency and low-frequency information in each branch separately. Extensive experiments show that our PointConT model achieves a remarkable performance on point cloud shape classification. Especially, our method exhibits 90.3% Top-1 accuracy on the hardest setting of ScanObjectN N. Source code of this paper is available at https://github.com/yahuiliu99/PointC onT.

Material point method simulation of hydro-mechanical behaviour in twophase porous geomaterials: A state-of-the-art review

The material point method(MPM)has been gaining increasing popularity as an appropriate approach to the solution of coupled hydro-mechanical problems involving large deformation.In this paper,we survey the current state-of-the-art in the MPM simulation of hydro-mechanical behaviour in two-phase porous geomaterials.The review covers the recent advances and developments in the MPM and their extensions to capture the coupled hydro-mechanical problems involving large deformations.The focus of this review is aiming at providing a clear picture of what has or has not been developed or implemented for simulating two-phase coupled large deformation problems,which will provide some direct reference for both practitioners and researchers.

Two-Staged Method for Ice Channel Identification Based on Image Segmentation and Corner Point Regression

Identification of the ice channel is the basic technology for developing intelligent ships in ice-covered waters,which is important to ensure the safety and economy of navigation.In the Arctic,merchant ships with low ice class often navigate in channels opened up by icebreakers.Navigation in the ice channel often depends on good maneuverability skills and abundant experience from the captain to a large extent.The ship may get stuck if steered into ice fields off the channel.Under this circumstance,it is very important to study how to identify the boundary lines of ice channels with a reliable method.In this paper,a two-staged ice channel identification method is developed based on image segmentation and corner point regression.The first stage employs the image segmentation method to extract channel regions.In the second stage,an intelligent corner regression network is proposed to extract the channel boundary lines from the channel region.A non-intelligent angle-based filtering and clustering method is proposed and compared with corner point regression network.The training and evaluation of the segmentation method and corner regression network are carried out on the synthetic and real ice channel dataset.The evaluation results show that the accuracy of the method using the corner point regression network in the second stage is achieved as high as 73.33%on the synthetic ice channel dataset and 70.66%on the real ice channel dataset,and the processing speed can reach up to 14.58frames per second.

Mobility and dynamic erosion process of granular flow:insights from numerical investigation using material point method

In order to understand the dynamics of granular flow on an erodible base soil,in this paper,a series of material point method-based granular column collapse tests were conducted to investigate numerically the mobility and dynamic erosion process of granular flow subjected to the complex settings,i.e.,the aspect ratio,granular mass,friction and dilatancy resistance,gravity and presence of water.A set of power scaling laws were proposed to describe the final deposit characteristics of granular flow by the relations of the normalized run-out distance and the normalized final height of granular flow against the aspect ratio,being greatly affected by the complex geological settings,e.g.,granular mass,the friction and dilatancy resistance of granular soil,and presence of water in granular flow.An index of the coefficient of friction of granular soil was defined as a ratio of the target coefficient of friction over the initial coefficient of friction to quantify the scaling extent of friction change(i.e.,friction strengthening or weakening).There is a characteristic aspect ratio of granular column corresponding to the maximum mobility of granular flow with the minimum index of the apparent coefficient of friction.The index of the repose coefficient of friction of granular flow decreased gradually with the increase in aspect ratio because higher potential energy of granular column at a larger aspect ratio causes a larger kinetic energy of granular soil to weaken the friction of granular soil as a kind of velocity-related friction weakening.An increase in granular mass reduces gradually the indexes of the apparent and repose coefficients of friction of granular soil to enhance the mobility of granular flow.The mobility of granular flow increases gradually with the decrease in friction angle or increase in dilatancy angle of granular soil.However,the increase of gravity accelerates granular flow but showing the same final deposit profile without any dependence on gravity.The mobility of granular flow increases gradually by lowering the indexes of the apparent and repose coefficients of friction of granular flow while changing the surroundings,in turn,the dry soil,submerged soil and saturated soil,implying a gradually increased excessive mobility of granular flow with the friction weakening of granular soil.Presence of water in granular flow may be a potential catalyzer to yield a long run-out granular flow,as revealed in comparison of water-absent and water-present granular flows.In addition,the dynamic erosion and entrainment of based soil induced by granular flow subjected to the complex geological settings,i.e.,the aspect ratio,granular mass,gravity,friction and dilatancy resistance,and presence of water,were comprehensively investigated as well.

Automatic Extraction Method of 3D Feature Guidelines for Complex Cultural Relic Surfaces Based on Point Cloud

Cultural relics line graphic serves as a crucial form of traditional artifact information documentation,which is a simple and intuitive product with low cost of displaying compared with 3D models.Dimensionality reduction is undoubtedly necessary for line drawings.However,most existing methods for artifact drawing rely on the principles of orthographic projection that always cannot avoid angle occlusion and data overlapping while the surface of cultural relics is complex.Therefore,conformal mapping was introduced as a dimensionality reduction way to compensate for the limitation of orthographic projection.Based on the given criteria for assessing surface complexity,this paper proposed a three-dimensional feature guideline extraction method for complex cultural relic surfaces.A 2D and 3D combined factor that measured the importance of points on describing surface features,vertex weight,was designed.Then the selection threshold for feature guideline extraction was determined based on the differences between vertex weight and shape index distributions.The feasibility and stability were verified through experiments conducted on real cultural relic surface data.Results demonstrated the ability of the method to address the challenges associated with the automatic generation of line drawings for complex surfaces.The extraction method and the obtained results will be useful for line graphic drawing,displaying and propaganda of cultural relics.

基于改进PointNet++的输电线路关键部位点云语义分割研究

输电线路的关键部位包括塔身、导线、绝缘子、避雷线以及引流线,无人机精细化导航的首要任务是构造输电线路的点云地图并从中分割出上述部位。为解决现有算法在输电线路的绝缘子、引流线等精细结构分割时精度低的问题,通过改进PointNet++算法,提出了一种面向输电线路精细结构的点云分割方法。首先,基于无人机机载激光雷达在现场采集的点云数据,构造了输电线路点云分割数据集;其次,通过对比实验,筛选出在本输电线路场景下合理的数据增强方法,并对数据集进行了数据增强;最后,将自注意力机制以及倒置残差结构和PointNet++相结合,设计了输电线路关键部位点云语义分割算法。实验结果表明:该改进PointNet++算法在全场景输电线路现场点云数据作为输入的前提下,首次实现了对引流线、绝缘子等输电线路中精细结构和导线、杆塔塔身以及输电线路无关背景点的同时分割,平均交并比(mean intersection over union,mIoU)达80.79%,所有类别分割的平均F_(1)值(F1 score)达88.99%。

Intelligent extraction of road cracks based on vehicle laser point cloud and panoramic sequence images

In light of the limited efficacy of conventional methods for identifying pavement cracks and the absence of comprehensive depth and location data in two-dimensional photographs,this study presents an intelligent strategy for extracting road cracks.This methodology involves the integration of laser point cloud data obtained from a vehicle-mounted system and a panoramic sequence of images.The study employs a vehicle-mounted LiDAR measurement system to acquire laser point cloud and panoramic sequence image data simultaneously.A convolutional neural network is utilized to extract cracks from the panoramic sequence image.The extracted sequence image is then aligned with the laser point cloud,enabling the assignment of RGB information to the vehicle-mounted three dimensional(3D)point cloud and location information to the two dimensional(2D)panoramic image.Additionally,a threshold value is set based on the crack elevation change to extract the aligned roadway point cloud.The three-dimensional data pertaining to the cracks can be acquired.The experimental findings demonstrate that the use of convolutional neural networks has yielded noteworthy outcomes in the extraction of road cracks.The utilization of point cloud and image alignment techniques enables the extraction of precise location data pertaining to road cracks.This approach exhibits superior accuracy when compared to conventional methods.Moreover,it facilitates rapid and accurate identification and localization of road cracks,thereby playing a crucial role in ensuring road maintenance and traffic safety.Consequently,this technique finds extensive application in the domains of intelligent transportation and urbanization development.The technology exhibits significant promise for use in the domains of intelligent transportation and city development.

Coexistence of Dirac and Weyl points in non-centrosymmetric semimetal NbIrTe_(4)

Using angle-resolved photoemission spectroscopy and density functional theory calculations methods,we investigate the electronic structures and topological properties of ternary tellurides NbIrTe_(4),a candidate for type-II Weyl semimetal.We demonstrate the presence of several Fermi arcs connecting their corresponding Weyl points on both termination surfaces of the topological material.Our analysis reveals the existence of Dirac points,in addition to Weyl points,giving both theoretical and experimental evidences of the coexistence of Dirac and Weyl points in a single material.These findings not only confirm NbIrTe_(4) as a unique topological semimetal but also open avenues for exploring novel electronic devices based on its coexisting Dirac and Weyl fermions.

PCT和PointMLP融合的三维点云分类算法研究

为了提高三维点云数据的分类准确率,对注意力机制的PCT与残差点MLP的PointMLP的融合算法进行研究。首先,加入几何仿射模块对输入点云进行正态分布以解决点云的密度不均匀和几何结构不确定性;其次,在编码模块首先使用注意力机制对上层特征进行全局表达,之后加入残差点MLP模块对注意力模块的输出特征进行深度特征表示以充分提取点云特征;最后是分类模块,完成下游点云分类任务。实验结果表明,改进后的算法在ModelNet40数据集上的分类准确率达到95.6%,相比PCT模型准确率提高了2.4%,且对点云数量的变化具有鲁棒性。

基于知识蒸馏和定位引导的Pointpillars点云检测网络

激光雷达数据由于其几何特性,被广泛应用于三维目标检测任务中。由于点云数据的稀疏性和不规则性,难以实现特征提取的质量和推理速度间的平衡。本文提出一种基于体柱特征编码的三维目标检测算法,以Pointpillars网络为基础,设计Teacher-Student模型框架对回归框尺度进行蒸馏,增加蒸馏损失,优化训练网络模型,提升特征提取的质量。为进一步提高模型检测效果,设计定位引导分类项,增加分类预测和回归预测之间的相关性,提高物体识别准确率。本网络所做改进没有引入额外的网络嵌入。算法在KITTI数据集上的实验结果表明,相比于基准网络,在三维模式下的平均精度值从60.65%提升到了64.69%,鸟瞰图模式下的平均精度值从67.74%提升到70.24%。模型推理速度为45 FPS,在提升检测精度的同时满足了实时性要求。