Author Affiliations
1School of Karst Science/School of Geography & Environmental Science, Guizhou Normal University, Guiyang 550001, Guizhou , China2State Engineering Technology Institute For for Karst Desertification Control, Guizhou Normal University, Guiyang 550001, Guizhou , Chinashow less
Fig. 1. Location diagram of study area. (a) DEM of Huajiang demonstration area; (b) visible remote sensing image of study area
Fig. 2. Method test area and profile of its point cloud data. (a) Visible remote sensing image; (b) image matching point cloud data;(c) point cloud data profile
Fig. 3. Schematic of extraction method of color index and CHM segmentation intersection fusion
Fig. 4. Technology roadmap
Fig. 5. Precision test standard of segmentation. (a) Match; (b) over matching; (c) over segmentation; (d) wrong segmentation; (e) missed inspection
Fig. 6. Histograms of number of pixels with different color indices. (a) VDVI; (b) NGBDI; (c) NGRDI; (d) RGBVI
Fig. 7. Extraction results of vegetation index. (a) VDVI; (b) NGBDI; (c) NGRDI; (d) RGBVI
Fig. 8. Preprocessing results of point cloud data. (a) DEM; (b) DSM; (c) CHM; (d) CHM segmentation
Fig. 9. Segmentation results of fusion of each color index and CHM segmentation. (a) VDVI_CHM; (b) NGBDI_CHM;(c) NGRDI_CHM; (d) RGBVI_CHM
Fig. 10. Comparison of extraction results between local fusion method and single factor extraction
Fig. 11. Box scatter diagram of four fusion results extraction and true value fitting degree
Fig. 12. Extraction process diagram and segmentation results of accuracy verification area. (a) DEM; (b) CHM; (c) VDVI; (d) VDVI_CHM
Training sample region | Red light band | Green light band | Blue light band |
---|
Pitaya area | 0.235 | 0.213 | 0.251 | Weeds | 0.191 | 0.154 | 0.189 | Bare land | 0.100 | 0.122 | 0.138 | Gravel ladder insuperior | 0.180 | 0.196 | 0.189 | Constructions | 0.126 | 0.122 | 0.107 |
|
Table 1. Variation coefficient of reflection value in visible light band of training sample area
Vegetation index | Equation |
---|
VDVI | | NGBDI | | NGRDI | | RGBVI | |
|
Table 2. Calculation formula of vegetation indexs
Segmentation type | PP | PC | SE | SC | L | P | R | F |
---|
VDVI | 59.38 | 34.06 | 0.00 | 6.56 | 0.00 | 59.38 | 39.83 | 47.68 | NGBDI | 61.00 | 37.67 | 0.00 | 1.33 | 0.63 | 61.00 | 38.36 | 47.10 | NGRDI | 69.72 | 21.88 | 0.00 | 8.40 | 4.19 | 69.72 | 57.44 | 62.99 | RGBVI | 58.69 | 39.02 | 0.00 | 2.30 | 0.63 | 58.69 | 37.53 | 45.78 |
|
Table 3. Accuracy table of pitaya tree number extracted by each vegetation index
Parameter | PP | PC | SE | SC | L | P | R | F |
---|
Value | 36.21 | 39.42 | 2.37 | 22.00 | 1.05 | 36.21 | 44.86 | 40.07 |
|
Table 4. Precision table of CHM segmentation
Segmentation type | PP | PC | SE | SC | L | P | R | F |
---|
VDVI_CHM | 94.06 | 1.70 | 3.18 | 1.06 | 1.05 | 94.06 | 92.87 | 93.46 | NGBDI_CHM | 93.05 | 2.95 | 2.74 | 1.26 | 1.05 | 93.05 | 92.66 | 92.86 | NGRDI_CHM | 93.60 | 1.77 | 2.21 | 2.43 | 6.92 | 93.60 | 88.89 | 91.18 | RGBVI_CHM | 92.26 | 3.35 | 1.26 | 3.14 | 1.68 | 92.26 | 92.45 | 92.36 |
|
Table 5. Accuracy table of extraction based on CHM segmentation and fusion of four color indices
Segmentation type | Increase in relative vegetation index | Increase relative to CHM segmentation |
---|
P | R | F | P | R | F |
---|
VDVI | 34.68 | 53.04 | 45.78 | 57.85 | 48.01 | 53.39 | NGBDI | 32.05 | 54.30 | 45.75 | 56.84 | 47.80 | 52.78 | NGRDI | 23.88 | 31.45 | 28.19 | 57.39 | 44.03 | 51.11 | RGBVI | 33.57 | 54.93 | 46.58 | 56.05 | 47.59 | 52.28 |
|
Table 6. Comparison table of segmentation fusion results and extraction accuracy of single factor
Parameter | Mean value | Standard deviation | Variance | Skewness | Peakedness |
---|
True_AREA | 133.800 | 37.496 | 1405.931 | 0.645 | 0.893 | VDVI_AREA | 144.293 | 48.244 | 2327.446 | 0.617 | 0.520 | RGBVI_AREA | 143.472 | 49.844 | 2484.471 | 0.862 | 1.874 | NGBDI_AREA | 143.571 | 51.018 | 2602.808 | 1.006 | 1.909 | NGRDI_AREA | 110.841 | 55.569 | 3087.924 | 1.502 | 3.984 |
|
Table 7. Description and analysis of basic data of real value and area value extracted by fusion