+ All Categories
Home > Documents > 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I...

1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I...

Date post: 21-Dec-2015
Category:
View: 215 times
Download: 1 times
Share this document with a friend
Popular Tags:
31
1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa Adams (Investigators) Nick Koncz (Project Manager) Hongwei Zhu, Amar Padmanabhan, Jisang Park (Research Assistants)
Transcript
Page 1: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

1Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy

Techniques

UW-Madison NCRST-I Research Team

Frank Scarpace, Alan Vonderohe, Teresa Adams (Investigators)

Nick Koncz (Project Manager)

Hongwei Zhu, Amar Padmanabhan, Jisang Park (Research Assistants)

Page 2: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

2Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy

Techniques

Objectives

Determine Differences among Results from the Various Techniques

Seek Methods for Improving Accuracies by Technology Integration

Seek Methods for Reducing Required Editing Time for Raw Softcopy Data

Page 3: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

Test Site: Highway Corridor Near Solon, IA

Page 4: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

4

Manual, LIDAR, and Softcopy Data Sets

Manual Photogrammetry Data Set Provided by Iowa DOT and CTRE:

Breaklines and Mass Points (~20-Meter Spacing)

Compiled on Analytical Stereoplotters from 1:4800 (nominal scale) photos

Expected Accuracy: 0.07-0.10m RMS

Page 5: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

5

Breaklines and Mass Points

Page 6: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

6

1-Meter DEM Generatedfrom Manual Photogrammetry Data Set

Page 7: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

7

Manual, LIDAR, and Softcopy Data Sets

Softcopy Photogrammetry Data Set:

Same Photography as Manual Method Same Camera Calibration Same External Orientation Parameters

Film Diapositives Scanned at 15 Micrometers

38 Photos in 3 Strips – 35 Stereo Models

Page 8: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

8

Manual, LIDAR, and Softcopy Data Sets

Softcopy Photogrammetry Data Set: In-House Software

Resampled Epipolar Images

1:32 Image Pyramids

Cross-Correlation

Least Squares Matching

Generates Irregular 1-Meter Spacing of Elevations

Page 9: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

Correlation Coefficients from a Single Model

Red = 0.5-0.7

Yellow = 0.7-0.9

Green = > 0.9

Page 10: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

DEM by Softcopy Photogrammetry

Page 11: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

11

Manual, LIDAR, and Softcopy Data Sets

LIDAR Data Set: Irregular 2-Meter Spacing of Elevations

Expected Accuracy: 0.15m RMS

Raw Data Were Edited, But Some Vegetation (e.g., Crops) Were Not Removed

Page 12: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

12

Part of the LIDAR Data Set

Page 13: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

13

Parts of the Three Data Sets

Sample Comparisons and Results

Page 14: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

Comparison Methodology

TINs

Manual PhotogrammetryMass Points and Breaklines

20m spacing (irregular)

LIDAR2 m spacing (irregular)

DigitalPhotogrammetry

1 m spacing (irregular)

POINTS

DEMs

DEMComparisons

POINT To TINComparisons

Page 15: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

Sample Comparisons and Results

DifferenceType

Mean RMS Mean RMS(m) (m) (m) (m)

Sub-Area 1: Farmland DEM 0.29 0.39 0.33 0.35Without crops PT-TIN 0.21 0.37 0.31 0.37

Sub-Area 2: Farms/Road DEM 0.29 0.33 -0.11 0.27Without crops PT-TIN -0.37 0.43 -0.21 0.33

Sub-Area 3: Mixed Use DEM -0.13 0.39 0.17 0.21Without Hwy PT-TIN -0.14 0.69 0.12 0.21

AVERAGE DEM 0.15 0.37 0.13 0.28PT-TIN -0.10 0.50 0.07 0.30

Location 1: Drainage Ditch DEM 0.14 0.59 0.09 0.31PT-TIN 0.22 0.57 0.11 0.41

Location 2: Gully DEM 0.00 0.43 0.09 0.26PT-TIN 0.23 0.64 0.13 0.35

Location 3: Bridge DEM -0.25 0.33 0.14 0.19Without bridge deck PT-TIN -0.17 0.39 0.14 0.24

AVERAGE DEM -0.04 0.45 0.11 0.25PT-TIN 0.09 0.53 0.13 0.33

Softcopy

Area+ 2 m Tolerance

with 1st Approx.

ExtractionLIDAR

Preliminary Results Indicate that Softcopy Data are at Least as Good as LIDAR when Compared to Manually-Extracted Data.

Page 16: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

16

Sample Comparisons and Results

Mixed Land Use

Page 17: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

17

Sample Comparisons and Results

Drainage Ditch

Page 18: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

18Softcopy / LIDAR Integration Project StatusSoftcopy Extraction w/LIDAR (Initial Comparison)

Model 127 PT-To-TIN Comparison Area #Points Min Max Mean StdDev RMS

Auto Match/ Manualw/o 1st approximation Whole DEM 3,615 -2.13 1.67 0.17 0.21 0.27

2m tolerance 3,615 -0.72 5.43 0.86 1.22 1.49

w/o 1st approximation W/O Road 1,202 -0.97 1.04 0.13 0.17 0.222m tolerance 1,202 -0.36 2.10 0.12 0.17 0.21

w/ LiDAR approximation + 1 m tolerance2.5M Spacing Whole DEM 3,615 -0.88 1.00 0.14 0.14 0.20

W/O Road 1,202 -0.74 1.00 0.10 0.15 0.18

5M Spacing Whole DEM 3,615 -0.87 0.97 0.14 0.13 0.19W/O Road 1,202 -0.63 0.97 0.08 0.14 0.16

10M Spacing Whole DEM 3,615 -1.19 0.75 0.13 0.14 0.19W/O Road 1,202 -0.63 0.71 0.08 0.14 0.16

LiDAR/ Manual Whole DEM 3,615 -1.59 2.89 -0.15 0.48 0.50W/O Road 1,202 -1.45 2.89 -0.14 0.68 0.69

LiDAR/ Auto Matchw/o 1st approximation Whole DEM 163,190 -5.55 2.99 -0.32 0.40 0.52

2m tolerance 134,705 -6.00 2.91 -0.33 0.40 0.52

w/o 1st approximation W/O Road 109,640 -2.67 2.99 -0.31 0.39 0.502m tolerance 104,124 -2.00 2.91 -0.31 0.34 0.46

Page 19: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

19Softcopy / LIDAR Integration Project Status

Softcopy Extraction w/LIDAR (Initial Comparison)

Drainage Ditch PT-To-TIN #Points Min Max Mean StdDev RMS

Auto Match/ Manual 215w/o 1st approximation -0.23 0.79 0.17 0.17 0.24

2m tolerance -3.36 1.04 0.11 0.39 0.41

w/ LiDAR approximation + 1 m tolerance2.5 m Spacing -0.55 0.90 0.18 0.18 0.25

5 m spacing -0.68 1.23 0.19 0.20 0.2810 m spacing -0.45 1.12 0.20 0.19 0.28

LiDAR/ Manual 215 -0.71 1.66 0.22 0.53 0.57

LiDAR/ Auto Matchw/o 1st approximation 14611 -1.45 1.93 -0.10 0.58 0.59

2m tolerance 93104 -1.91 4.23 -0.06 0.58 0.58

Page 20: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

20

Softcopy Editing Tools

Automated Slope Filter (Spikes and Holes)

Manual (Stereo Viewing) Point-by-Point

Polygon Constant Elevation

Polygon Planar Fit

Page 21: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

21

Manual Editing Tool Menu

Page 22: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

22

Manual Editing Polygon Selection Tool

Page 23: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

23

Manual Editing Set-to-Constant Elevation Tool

Page 24: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

One of the Stereo Pairs

Page 25: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

Raw Softcopy Data

Page 26: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

After Slope Filter

Page 27: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

After Slope Filter

Page 28: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

After Manual Editing

Page 29: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

After Manual Editing

Page 30: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

Effects of Slope Filter(Crop area was excluded)

Model: 4-1-125

Difference DEM Area #Points Min Max Mean StdDev RMS

Auto Match/ Manualw/ USGS approximation

Whole DEM 5044 -3.88 1.66 -0.19 0.5 0.54

W/O Road 3793 -3.88 1.66 -0.16 0.5 0.52

w/ LiDAR approximationWhole DEM 5044 -2.23 3.61 -0.11 0.46 0.47(without filter)W/O Road 3793 -2.23 3.61 -0.07 0.46 0.47(without filter)Whole DEM 5044 -2.16 1.84 -0.11 0.41 0.43(with filter)W/O Road 3793 -2.16 1.84 -0.1 0.41 0.43(with filter)

Page 31: 1 Comparison of Surface Models Derived by Manual, LIDAR, and Softcopy Techniques UW-Madison NCRST-I Research Team Frank Scarpace, Alan Vonderohe, Teresa.

31

ConclusionsWhen Differenced with Manually-Derived Data, Softcopy Results (0.2-0.4m RMS) are Slightly Better than LIDAR (0.3-0.5m RMS).When LIDAR is Used as First Approximation for Softcopy, Results are Mixed with Improvements of 20% (to 0.16m RMS) in Some Cases.Slope Filter Improves Raw Softcopy Data by 10%.Comparisons with Manually-Edited Softcopy Remain to be Done.


Recommended