things to keep in mind while scanning get excellent data to ......things to keep in mind while...
TRANSCRIPT
Things to keep in mind while scanning-
Get excellent data to work with
Richard SteffenCTO
Martin GranerR&D Engineer
Things to keep in mind - 2020
Table of contents
1. Project planning2. Scanning3. Registration
2
Things to keep in mind - 2020
Table of contents
1. Project planninga. Requirementsb. Time x moneyc. Planning
3
Things to keep in mind - 2020
What does the customer want / need:
● Accuracy and point density● Point cloud for presentation● Processing for BIM / CAD
Project planning - Requirements
4
Things to keep in mind - 2020
On accuracy:
● Global vs local accuracy● Verification
Project planning - Requirements
5
Things to keep in mind - 2020
One scan
Project planning - Requirements
Two scans
6
Things to keep in mind - 2020
● Scanning with / without colour● Resolution● Estimated Scan positions● Data accuracy - mandatory things to achieve it
Project planning - Time x money
7
TODO: Übersichtstabelle mit Scanning zeiten je auflösung…
https://docs.google.com/spreadsheets/d/1mJyidKomhSUzHRKro8PFIrJw-IO1Gq7W9At6vDOW3q4/edit#gid=1535648332
Project planning - Time x money
8
TODO: Übersichtstabelle mit Scanning zeiten je auflösung…
https://docs.google.com/spreadsheets/d/1mJyidKomhSUzHRKro8PFIrJw-IO1Gq7W9At6vDOW3q4/edit#gid=1535648332
Project planning - Time x money
9
TODO: Übersichtstabelle mit Scanning zeiten je auflösung…
https://docs.google.com/spreadsheets/d/1mJyidKomhSUzHRKro8PFIrJw-IO1Gq7W9At6vDOW3q4/edit#gid=1535648332
Project planning - Time x money
10
TODO: Übersichtstabelle mit Scanning zeiten je auflösung…
https://docs.google.com/spreadsheets/d/1mJyidKomhSUzHRKro8PFIrJw-IO1Gq7W9At6vDOW3q4/edit#gid=1535648332
Project planning - Time x money
11
Things to keep in mind - 2020
Project planning - Time x money
12
Beam Divergence Spot Size
Distance ----------------------------------------------- >
Things to keep in mind - 2020
Project planning - Time x money
13
Things to keep in mind - 2020
Project planning - Time x money
14
Single Point Accuracy
Distance ----------------------------------------------- >
ςα
ςd
ςα = const (Faro 19 arcsec = 0.005°)
ςd = const + ppm (Faro 1mm + 10 ppm)
Things to keep in mind - 2020
Project planning - Time x money
15
Things to keep in mind - 2020
Why beam divergence and single point accuracy is important?
● Beam divergence○ Reflected waveform has multiple peaks in time diagramm○ Usually mean will be computed (ghost points on edges)
● Single point Accuracy○ Should be considered in the adjustment○ Should be considered in distance measurements
Project planning - Planning
16
Things to keep in mind - 2020
● Use existing Layout plan! (Scan position planning & field book)● Loop closures, feature placements● Geodetics● Use high standard spheres and checkerboard targets● Do a site inspection prior to scanning● On-site registration
Project planning - Planning
17
Project planning - Planning
18
19
Things to keep in mind - 2020
Table of contents
1. Project planning2. Scanning
a. C2C vs feature basedb. Georeferencingc. How to scan
20
Things to keep in mind - 2020
Scanning - C2C vs feature based
21
C2C Feature based
Matching via: ● Point to Point (ICP)● Point to Plane (ICP*)● Planes (patches)
● Spheres● Checkerboard targets● Natural features
Limits: ● No planes/normals in three axis existing (e.g. long hallways)
● Not enough features (<3)● Bad feature constellation
Gets time consuming when:
● No approximation values (Creating the connection graph)Manual Topview pre-alignment
● Bad Automatic feature extraction● Getting wrong approximation
values
Constellation search: ● Scan to scan connections ● Global Scan connections
Things to keep in mind - 2020
● Is it necessary?● Direct vs indirect● While registering vs after registration
Scanning - Georeferencing
22
Things to keep in mind - 2020
● Scan position connections● Loop closure● Feature distribution● How to scan through a door● Automatic sphere detection
Scanning - How to scan
23
Things to keep in mind - 2020
Scanning - How to scan
24
Scan:
Scan connection
Things to keep in mind - 2020
Scanning - How to scan
25
Scan:
Scan connection
Things to keep in mind - 2020
Scanning - How to scan
26
Loop closure
Scan:
Things to keep in mind - 2020
Scanning - How to scan
27
Loop closure
Scan:
Things to keep in mind - 2020
Scanning - How to scan
28
Feature distribution
Feature:
Scan:
Things to keep in mind - 2020
Scanning - How to scan
29
The good
Feature:
Scan:
Things to keep in mind - 2020
Scanning - How to scan
30
The bad
Feature:
Scan:
Things to keep in mind - 2020
Scanning - How to scan
31
The ugly
Feature:
Scan:
Things to keep in mind - 2020
Scanning - How to scan
32
The super bad luck
60°
60° 60°
Feature:
Scan:
Things to keep in mind - 2020
Scanning - How to scan
33
The super bad luck
Feature:
Scan:
Things to keep in mind - 2020
Scanning - How to scan
34
The super bad luck
Feature:
Scan:
Things to keep in mind - 2020
Scanning - How to scan
35
The super bad luck
Feature:
Scan:
Things to keep in mind - 2020
Scanning - How to scan
36
Door scanningConfiguration 1
Things to keep in mind - 2020
Scanning - How to scan
37
Door scanningConfiguration 2
Scanning - How to scan
38
Source: https://www.laserscanning-europe.com/en/news/use-laser-scanner-reference-spheres-optimal-distance-to-the-scanner
39
Real life sphere detection in Faro Scene
Things to keep in mind - 2020
Table of contents
1. Project planning2. Scanning3. Registration
a. Feature basedb. Cloud to Cloud (C2C)
40
Things to keep in mind - 2020
The big challenge
● Determine the connection graph○ Which Scans are overlapped (adjacency matrix)○ Which local features correspond to each other (constellation search)
● Simplify automatic search algorithms○ Complexity O (n2 / 2)○ Clustering 20->100 Scans per cluster
Registration
41
Things to keep in mind - 2020
● Wording issues: “Bundle adjustment” - It’s Photogrammetry, not scanning because of the 3D->2D Projection (bundle of rays)
● Least squares adjustment -> Squared sum of the residuals (errors) will be minimized
Registration
42
Things to keep in mind - 2020
Feature based model
● Each Scan depends on H (X, Y, Z, ⍵, φ, ϰ) - rigid motion● Model for point based features: Xw = Hw Xlocal
○ Xw Unknown 3D Position○ Xlocal Measured local 3D Position
● Georefernce Observation○ Xw = Xgeod
● Inclinometer Observation⍵ = ⍵inc , ϰ = ϰinc
Registration - Feature based
43
Things to keep in mind - 2020
Global Adjustment - Feature Based
● All “raw” observations with their accuracy introduced in one big adjustment (usually sparse system)
● All errors are distributed into the whole network● For all observation types a posteriori accuracy can be estimated● Full error analyze possible● Georeferencing included, reduce drift effect
Registration
44
Things to keep in mind - 2020
“Two step” alignment - Used very often for Cloud2Cloud
● 1. Step: Alignment of relative orientation between two scans○ The model: X2 = ΔH12 X1 (Helmert transformation between 2 scans)○ Cloud 2 Cloud use IPC to detemine ΔH12
● 2. Step: Net adjustment of all relative orientations in a cluster or for complete overlapping graph
○ The model: Hw2 = ΔH12 Hw1 (Relative orientation)
● 3. Step: Cluster alignment (big residual stress on cluster bounds)● 4. Step: Georeferencing (big residual stress)
Registration - Cloud to Cloud
45
Things to keep in mind - 2020
Recommendations for the registration
1. Use Cloud2Cloud for small projects (20-50 scan) and no georeferencing
2. Use Feature based Registration for proof of result
3. Introduce high accuracy control points (GPS/Total station) for Georeferencing and drift compensation
Registration - Cloud to Cloud
46