建筑物立面点云直线段特征提取方法 下载: 1096次
Straight-Line-Segment Feature-Extraction Method for Building-Facade Point-Cloud Data
1 同济大学测绘与地理信息学院, 上海 200092
2 自然资源部现代工程测量重点实验室, 上海 200092
图 & 表
图 1. 原始点云姿态调整。(a)原始点云;(b)姿态调整后点云
Fig. 1. Attitude adjustment of original point cloud. (a) Original point cloud; (b) point cloud after attitude adjustment
下载图片 查看原文
图 2. 姿态调整。(a)原始点云姿态;(b)原始点云主方向;(c)调整后点云姿态
Fig. 2. Attitude adjustment. (a) Attitude of original point cloud; (b) principal direction of original point cloud; (c) attitude of point cloud after adjustment
下载图片 查看原文
图 3. 切片生成。(a) X轴切片;(b) Y轴切片;(c) Z轴切片
Fig. 3. Slice generation. (a) Slices along X axis; (b) slices along Y axis; (c) slices along Z axis
下载图片 查看原文
图 4. 切片上特征点提取
Fig. 4. Extraction of feature points from slices
下载图片 查看原文
图 5. 近拐角点
Fig. 5. Points near corner
下载图片 查看原文
图 6. 映射回三维空间的特征点
Fig. 6. Feature points mapped back into three-dimensional space
下载图片 查看原文
图 7. 特征点提取。(a) X轴切片特征点;(b) Y轴切片特征点;(c) Z轴切片特征点
Fig. 7. Extraction of feature points. (a) Feature points on slices along X axis; (b) feature points on slices along Y axis; (c) feature points on slices along Z axis
下载图片 查看原文
图 8. 基于圆柱体生长的特征点聚类流程图(Z方向)
Fig. 8. Flow chart of feature points clustered based on cylindrical growth (Z direction)
下载图片 查看原文
图 9. 直线段生长示意图
Fig. 9. Schematic of straight-line-segment growth
下载图片 查看原文
图 10. 特征点聚类。 (a) X轴切片特征点聚类;(b) Y轴切片特征点聚类;(c) Z轴切片特征点聚类
Fig. 10. Clustering of feature points. (a) Clustering of feature points on slices along X axis; (b) clustering of feature points on slices along Y axis; (c) clustering of feature points on slices along Z axis
下载图片 查看原文
图 11. 拟合的直线段
Fig. 11. Fitted straight line segments
下载图片 查看原文
图 12. 直线段端点调整与优化。(a)优化与调整前;(b)优化与调整后
Fig. 12. Adjustment and refinement of endpoints of straight line segments. (a) Before adjustment and refinement; (b) after adjustment and refinement
下载图片 查看原文
图 13. 直线段提取结果。(a)~(d) Data 1,Data 2,Data 3,Data 4;(e)~(h) Data 1,Data 2,Data 3,Data 4的提取结果
Fig. 13. Extraction results of straight line segments. (a)-(d) Data 1, Data 2, Data 3, and Data 4; (e)-(h) extraction results of Data 1, Data 2, Data 3, and Data 4
下载图片 查看原文
图 14. 直线段提取结果对比。(a) Data 1,文献[
9]算法提取结果;(b) Data 1,本文算法提取结果;(c) Data 3,文献[
9]算法提取结果;(d) Data 3,本文算法提取结果
Fig. 14. Comparison of extraction results of straight line segments. (a) Extraction result of Data 1 obtained by method in Ref. [9];(b) extraction result of Data 1 obtained by proposed method; (c) extraction result of Data 3 obtained by method in Ref. [9]; (d) extraction result of Data 3 obtained by proposed method
下载图片 查看原文
图 15. 精确率和召回率统计区域。(a) Data 1统计区域;(b) Data 3统计区域
Fig. 15. Statistical areas of precision and recall. (a) Statistical areas of Data 1; (b) statistical areas of Data 3
下载图片 查看原文
图 16. Data 1特征点提取结果对比。(a) PCA提取结果,0.2 cm<R<7 cm;(b) PCA提取结果,0.2 cm<R<10 cm;(c) PCA提取结果,0.2 cm<R<15 cm;(d)本文算法提取结果
Fig. 16. Comparison of extraction results of feature points of Data 1. (a) Extraction result of PCA, 0.2 cm<R<7 cm;(b) extraction result of PCA, 0.2 cm<R<10 cm; (c) extraction result of PCA, 0.2 cm<R<15 cm;(d) extraction result of proposed method
下载图片 查看原文
表 1实验数据处理结果
Table1. Processing results of experiment data
Data | Pointnumber | Pointinterval /cm | R /cm | Time cost /s |
---|
Slicing andFP extracting | FP clusteringand fitting | Post-processing | Total |
---|
Data 1 | 1776255 | 1.01 | 3.2 | 70.6 | 14.8 | 2.4 | 87.8 | Data 2 | 459970 | 3.14 | 7 | 30.1 | 8.3 | 1.9 | 40.3 | Data 3 | 325338 | 5.64 | 10 | 25.2 | 6.6 | 0.8 | 32.6 | Data 4 | 2338424 | 0.98 | 78 | 68.9 | 4 | 0.3 | 73.2 |
|
查看原文
表 2Data 4窗体提取误差分析
Table2. Error analysis of windows extraction of Data 4
Aspect | Designedvalue | Minimumerror | Maximumerror | Average |
---|
Width /mm | 900 | 2.9 | 18.4 | 9.15 | Height /mm | 1500 | 0 | 18.2 | 9.21 |
|
查看原文
表 3直线段提取结果的精确率和召回率对比
Table3. Comparison of precision and recall of extraction results of straight line segments
Parameter | Method | Data 1 | Data 3 | Average |
---|
1 | | 2 | 3 | 4 | 5 | 6 | 7 | 8 |
---|
Rprecision /% | Method in Ref.[9] | 90.5 | 86.7 | 100 | 93.8 | 85.0 | 100 | 88.2 | 85.7 | 91.2 | | Method in this paper | 91.1 | 91.8 | 100 | 91.7 | 83.3 | 100 | 96.2 | 94.4 | 93.6 | Rrecall /% | Method in Ref.[9] | 38.8 | 21.7 | 37.5 | 36.6 | 51.5 | 60.6 | 65.2 | 42.9 | 44.4 | | Method in this paper | 83.7 | 96.6 | 87.5 | 94.3 | 95.2 | 97.3 | 96.2 | 89.5 | 92.5 |
|
查看原文
李金涛, 程效军. 建筑物立面点云直线段特征提取方法[J]. 中国激光, 2019, 46(11): 1109002. Jintao Li, Xiaojun Cheng. Straight-Line-Segment Feature-Extraction Method for Building-Facade Point-Cloud Data[J]. Chinese Journal of Lasers, 2019, 46(11): 1109002.