In view of the limitations of traditional measurement methods in the field of building information,such as complex operation,low timeliness and poor accuracy,a new way of combining three-dimensional scanning technolog...In view of the limitations of traditional measurement methods in the field of building information,such as complex operation,low timeliness and poor accuracy,a new way of combining three-dimensional scanning technology and BIM(Building Information Modeling)model was discussed.Focused on the efficient acquisition of building geometric information using the fast-developing 3D point cloud technology,an improved deep learning-based 3D point cloud recognition method was proposed.The method optimised the network structure based on RandLA-Net to adapt to the large-scale point cloud processing requirements,while the semantic and instance features of the point cloud were integrated to significantly improve the recognition accuracy and provide a precise basis for BIM model remodeling.In addition,a visual BIM model generation system was developed,which systematically transformed the point cloud recognition results into BIM component parameters,automatically constructed BIM models,and promoted the open sharing and secondary development of models.The research results not only effectively promote the automation process of converting 3D point cloud data to refined BIM models,but also provide important technical support for promoting building informatisation and accelerating the construction of smart cities,showing a wide range of application potential and practical value.展开更多
In view of the large amount of data and dense pixel points in point cloud files,this article proposes a multiple point cloud file encryption algorithm based on principal component analysis(PCA)and fractional Fourier t...In view of the large amount of data and dense pixel points in point cloud files,this article proposes a multiple point cloud file encryption algorithm based on principal component analysis(PCA)and fractional Fourier transform(FrFT).In this method,a point cloud data matrix(PCDM)is generated by extracting the coordinates and color information of the point cloud,then using PCA to reduce the dimension of a sequence of PCDMs,which are spliced and scrambled to produce a feature vector matrix and a dimension-reduced matrix(DRM)for encryption and reconstruction.Then using the hyperchaotic Lorenz system to generate the random phase masks and the orders of the FrFT.These two parameters will be used as keys to encrypt the point cloud feature vector matrix.The simulation results verify that the encryption algorithm can quickly encrypt multiple point cloud files,and the quality of the point cloud files obtained by decryption and reconstruction is good.The algorithm also has a large enough key space and highly sensitive keys,which means it has good security and strong robustness to different attacks.展开更多
针对自动驾驶场景下,近处干扰点云误检率高、远处稀疏点云漏检率高的问题,提出了一种基于改进PointPillars的自动驾驶障碍物点云检测算法.首先,通过聚合模块和共享多层感知机(shared multi-layer perceptron,MLP)对柱体内点云进行特征编...针对自动驾驶场景下,近处干扰点云误检率高、远处稀疏点云漏检率高的问题,提出了一种基于改进PointPillars的自动驾驶障碍物点云检测算法.首先,通过聚合模块和共享多层感知机(shared multi-layer perceptron,MLP)对柱体内点云进行特征编码,采用最大池化与平均池化叠加的方法将点云的显著特征与细节特征映射为柱体特征;其次,针对算法对伪图特征关注与利用不充分的问题,引入坐标注意力(coordinate attention,CA)机制和残差连接的伪图特征提取模块(attention and residual second block,ARSB),将深层与浅层特征图进行融合,优化算法梯度,增强算法对有效目标的关注度.试验结果表明:改进算法对全局点云检测精度较高,平均精度优于PointPillars、稀疏到稠密3D目标检测器(STD)等点云目标检测算法,在汽车类别上的检测精度优势明显,检测速度较快,符合实时性要求.展开更多
针对菇房内杏鲍菇表型参数测量任务中,由于扫描设备视角受限,扫描的杏鲍菇点云出现残缺问题,基于AdaPoinTr(Adaptive geometry-aware point transformers)提出了改进的SwinPoinTr模型,实现了对残缺杏鲍菇点云的准确补全和杏鲍菇表型参...针对菇房内杏鲍菇表型参数测量任务中,由于扫描设备视角受限,扫描的杏鲍菇点云出现残缺问题,基于AdaPoinTr(Adaptive geometry-aware point transformers)提出了改进的SwinPoinTr模型,实现了对残缺杏鲍菇点云的准确补全和杏鲍菇表型参数的测量。该方法在使用提出的特征重塑模块的基础上,构建具有几何感知能力的层次化Transformer编码模块,提高了模型对输入点云的利用率和模型捕捉点云细节特征的能力。然后基于泊松重建方法完成了补全点云表面重建,并测量到杏鲍菇表型参数。实验结果表明,本文所提算法在残缺杏鲍菇点云补全任务中,模型倒角距离为1.316×10^(-4),地球移动距离为21.3282,F1分数为87.87%。在表型参数估测任务中,模型对杏鲍菇菌高、体积、表面积估测结果的决定系数分别为0.9582、0.9596、0.9605,均方根误差分别为4.4213 mm、10.8185 cm^(3)、7.5778 cm^(2)。结果证实了该研究方法可以有效地补全残缺的杏鲍菇点云,可以为菇房内杏鲍菇表型参数测量提供基础。展开更多
针对奶绵羊三维重构中背景分割对复杂场景适应性不足、配准算法对初始位置敏感等问题,该研究提出一种融合改进PointNet++与一致性点漂移(coherent point drift,CPD)算法与局部区域重叠的三维重构方法。通过引入点对特征、优化采样策略...针对奶绵羊三维重构中背景分割对复杂场景适应性不足、配准算法对初始位置敏感等问题,该研究提出一种融合改进PointNet++与一致性点漂移(coherent point drift,CPD)算法与局部区域重叠的三维重构方法。通过引入点对特征、优化采样策略及损失函数,增强了PointNet++在复杂场景下的分割能力;结合CPD算法与局部区域重叠策略,提升了点云配准的鲁棒性和效率。试验结果显示:该方法用于奶绵羊背景分割的准确率和平均交并比分别达到98.78%和97.25%,推理速度为53.4 ms;较原模型平均准确率和平均交并比分别提高了3.04和2.53个百分点,推理时间缩短了45.17%。该方法用于奶绵羊三维配准中,各向异性旋转误差、各向异性平移误差、各向同性旋转误差、各向同性平移误差以及倒角距离分别达到0.0256°、0.0229 m、3.0887°、0.0463 m和0.00789 m,较原始CPD方法均降低。通过与人工体尺测量数据对比,重构模型所提取的体长、体高、十字部高、胸深、胸围等参数的平均绝对百分比误差分别为3.34%、3.07%、3.32%、3.63%和2.81%。该研究方法兼具较高精度与实时性,能够满足一次性重构的需求,可为奶绵羊三维配准与智能化体尺测定提供参考。展开更多
文摘In view of the limitations of traditional measurement methods in the field of building information,such as complex operation,low timeliness and poor accuracy,a new way of combining three-dimensional scanning technology and BIM(Building Information Modeling)model was discussed.Focused on the efficient acquisition of building geometric information using the fast-developing 3D point cloud technology,an improved deep learning-based 3D point cloud recognition method was proposed.The method optimised the network structure based on RandLA-Net to adapt to the large-scale point cloud processing requirements,while the semantic and instance features of the point cloud were integrated to significantly improve the recognition accuracy and provide a precise basis for BIM model remodeling.In addition,a visual BIM model generation system was developed,which systematically transformed the point cloud recognition results into BIM component parameters,automatically constructed BIM models,and promoted the open sharing and secondary development of models.The research results not only effectively promote the automation process of converting 3D point cloud data to refined BIM models,but also provide important technical support for promoting building informatisation and accelerating the construction of smart cities,showing a wide range of application potential and practical value.
基金supported by National Natural Science Foundation of China(61771155).
文摘In view of the large amount of data and dense pixel points in point cloud files,this article proposes a multiple point cloud file encryption algorithm based on principal component analysis(PCA)and fractional Fourier transform(FrFT).In this method,a point cloud data matrix(PCDM)is generated by extracting the coordinates and color information of the point cloud,then using PCA to reduce the dimension of a sequence of PCDMs,which are spliced and scrambled to produce a feature vector matrix and a dimension-reduced matrix(DRM)for encryption and reconstruction.Then using the hyperchaotic Lorenz system to generate the random phase masks and the orders of the FrFT.These two parameters will be used as keys to encrypt the point cloud feature vector matrix.The simulation results verify that the encryption algorithm can quickly encrypt multiple point cloud files,and the quality of the point cloud files obtained by decryption and reconstruction is good.The algorithm also has a large enough key space and highly sensitive keys,which means it has good security and strong robustness to different attacks.
文摘针对自动驾驶场景下,近处干扰点云误检率高、远处稀疏点云漏检率高的问题,提出了一种基于改进PointPillars的自动驾驶障碍物点云检测算法.首先,通过聚合模块和共享多层感知机(shared multi-layer perceptron,MLP)对柱体内点云进行特征编码,采用最大池化与平均池化叠加的方法将点云的显著特征与细节特征映射为柱体特征;其次,针对算法对伪图特征关注与利用不充分的问题,引入坐标注意力(coordinate attention,CA)机制和残差连接的伪图特征提取模块(attention and residual second block,ARSB),将深层与浅层特征图进行融合,优化算法梯度,增强算法对有效目标的关注度.试验结果表明:改进算法对全局点云检测精度较高,平均精度优于PointPillars、稀疏到稠密3D目标检测器(STD)等点云目标检测算法,在汽车类别上的检测精度优势明显,检测速度较快,符合实时性要求.
