Infrared and Laser Engineering, Volume. 50, Issue 10, 20200501(2021)
Application of Gaussian Mixture Clustering to moving surface fitting filter classification
Figures & Tables(13)
Fig. 2. Point cloud real elevation and fitting elevation distribution
Fig. 8. Three types of errors and Kappa coefficient distributions of the two types of clustering algorithms
Fig. 9. Two types of clustering algorithms show the data processing of Samp53
Fig. 10. Two types of clustering algorithms show the processing of the specified grid data in Samp53
Table 1. Three types of errors and Kappa coefficient judgment method
View tableView in ArticleTable 1. Three types of errors and Kappa coefficient judgment method
Test sample Sample name Topographic features Site1 Samp11 Vegetation and buildings on steep slopes Samp12 Small features Site2 Samp21 Narrow bridge Samp22 Bridge/passage Samp23 Complex and huge buildings, intermittent terrain Samp24 Slope Site3 Samp31 Cluster low-value points (multi-path effect) Site4 Samp41 Discontinuous terrain Samp42 Buildings, high-frequency undulating features Site5 Samp51 Vegetation on the slope Samp52 Low vegetation, fractured steep ridges Samp53 Intermittent terrain Samp54 Low-resolution building points Site6 Samp61 Low-resolution building points Site7 Samp71 Bridge, discontinuous terrain
Table 2. Topographic features and error distribution of the survey area
View tableView in ArticleTable 2. Topographic features and error distribution of the survey area
Total number Type I error Type II error Total error PA PC Kappa pe=pa+pb+pc+pd pb/(pa+pb) pc/(pc+pd) (pb+pc)/pe (pa+pd)/pe [(pa+pb)(pa+pc)+(pc+pd)(pb+pd)]/pe2 (PA-PC)/(1-PC)
Table 3. Two types of clustering algorithms for data sample error statistics
View tableView in ArticleTable 3. Two types of clustering algorithms for data sample error statistics
Algorithm type Gaussian clustering model classification algorithm Pedigree clustering model classification algorithm Sample name Type I error Type II error Total error Kappa coefficient Type I error Type II error Total error Kappa coefficient Samp11 3.43% 14.38% 5.40% 81.79% 22.25% 6.38% 13.82% 72.00% Samp12 3.68% 6.25% 5.15% 89.56% 10.06% 6.01% 8.36% 83.01% Samp21 1.59% 4.52% 2.54% 94.19% 8.89% 8.64% 8.81% 80.37% Samp22 7.31% 6.15% 6.48% 84.16% 10.81% 22.41% 13.09% 61.71% Samp23 9.16% 9.81% 9.56% 80.13% 15.38% 16.82% 16.26% 66.47% Samp24 3.70% 14.2% 7.32% 83.45% 28.82% 10.15% 20.12% 60.15% Samp31 4.04% 6.24% 5.22% 89.51% 9.18% 12.50% 10.00% 74.48% Samp41 1.95% 2.17% 2.04% 95.81% 5.61% 3.53% 4.77% 90.18% Samp42 5.56% 3.87% 4.21% 87.53% 7.42% 7.46% 7.45% 84.46% Samp51 4.91% 4.84% 4.90% 86.23% 16.39% 8.51% 12.96% 74.01% Samp52 9.36% 10.53% 9.72% 77.84% 14.67% 7.27% 11.54% 76.76% Samp53 5.75% 15.79% 6.48% 60.69% 10.27% 34.04% 14.39% 52.59% Samp54 5.34% 1.30% 3.16% 93.62% 5.13% 12.23% 8.98% 82.03% Samp61 2.67% 5.88% 2.85% 76.53% 10.94% 0.00% 9.15% 72.69% Samp71 3.23% 7.25% 4.26% 88.60% 5.29% 12.90% 7.80% 82.26%
Tools
Get Citation
Copy Citation Text
Chengbin Xing, Shengsheng Gong, Xiaoliang Yu, Yixin Li. Application of Gaussian Mixture Clustering to moving surface fitting filter classification[J]. Infrared and Laser Engineering, 2021, 50(10): 20200501
Paper Information
Category: Image processing
Received: Dec. 18, 2020
Accepted: --
Published Online: Dec. 7, 2021
The Author Email: Shengsheng Gong (csu_gong@sina.com.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20