The Deviation Analysis Method for Quality Assessment of As-is BIMs

Assessment of Quality of As-is Building Information Models Generated from Point Clouds Using Deviation Analysis :

Assessment of Quality of As-is Building Information Models Generated from Point Clouds Using Deviation Analysis Engin Burak Anil, Pingbo Tang, Burcu Akinci , Daniel Huber

As-is Building Information Models (BIMs):

2 As-is Building Information Models (BIMs)

How As-is BIMs are Created:

How As-is BIMs are Created 3 Data collection Registration Modeling

The Physical Measurement Method for QA in As-is BIMs:

4 The Physical Measurement Method for QA in As-is BIMs Take physical measurements Conduct statistical analysis

The Deviation Analysis Method:

The Deviation Analysis Method Find differences between as-is BIM and raw point clouds or Find differences between point clouds from different scans Analyze differences for patterns of deviations 5 Deviation Analysis Deviation Patterns

The Deviation Analysis Method:

The Deviation Analysis Method Overlay BIM and point cloud Decide on tolerance values Decide on type of analysis and visualization style Perform analysis Analyze deviation patterns 6

Visualization Methods:

Visualization Methods 7 Continuous signed Continuous unsigned

Visualization Methods:

Visualization Methods 8 Binary map Histogram

Benefits of the Deviation Analysis Method:

Benefits of the Deviation Analysis Method Full coverage – Apply to any surface Ability to pinpoint error source – Different phases give different patterns Potential for automation – Can use computer vision techniques to detect patterns No need for additional physical access – Just use initial data Intermediate results – Can analyze raw data right away Can assess entire site – Limited only by visibility of surfaces 9

Identifying Errors: Data Collection Phase:

Identifying Errors: Data Collection Phase Calibration error Mixed pixel effect Specular reflections Moving objects 10

Identifying Errors: Data Registration Phase:

Identifying Errors: Data Registration Phase Alignment error patterns 11 Rotation error (synthetic) Translation error (synthetic) Single scan alignment error (real)

Identifying Errors: Modeling Phase:

2 3 1 1 1 Identifying Errors: Modeling Phase Types of errors Missing components Incorrect geometry Incorrect positioning Incorrect component type 12 Region 1: Missing windows in the model Region 2: Incorrectly positioned indentation on wall Region 3: Incorrect door location

Relationship to Previous Work:

Relationship to Previous Work 13 θ = 89.44 ° QA in manufacturing Clash detection Construction site inspection

Can Current Software Tools Do Deviation Analysis?:

Can Current Software Tools Do Deviation Analysis? Conducted a survey and evaluation of capabilities of existing tools Identified existing point cloud processing software (29 found) Narrowed to 5 software packages Performed an in-depth evaluation on these 14

Evaluation Criteria:

Evaluation Criteria Deviation measurement and visualization – points-to-points, points-to-BIM Interoperability – import BIM, import point cloud Performance on large data sets – millions to billions of points 15

Evaluation Results: Measurement and Visualization:

Evaluation Results: Measurement and Visualization Surface coloring method does not work as expected Working in the interiors is difficult 16 Point coloring Surface coloring

Evaluation Results: Interoperability:

Evaluation Results: Interoperability 17 Cannot import BIM formats directly Surface normal reversing Point cloud altering problems Surface orientation Corresponding deviation map

Evaluation Results: Large Data Sets:

Evaluation Results: Large Data Sets 18 Software was not able to handle practical data sets Dividing the data into interest regions was necessary

The Deviation Analysis Process in Practice:

Compute Deviation Map Overlay Point Cloud and BIM Data For Each Surface of Interest Extract Point Cloud Extract BIM Data Analyze Deviations 2 3 1 1 1 The Deviation Analysis Process in Practice 19 2 3 1 1 1

Comparison to the Physical Measurement Method:

Comparison to the Physical Measurement Method NIST/CMU case study 285 measurements using laser distance meter or tape We analyzed the using the deviation analysis method Compared on two dimensions: Agreement of measurements Coverage 20

Agreement of Measurements:

Agreement of Measurements 21 Among 24 physical measurements 9 were taken at non-existing components 14 out of the remaining 15 measurements agreed within 1 cm

Coverage:

Coverage 22 What does “coverage” mean with sparse sampling? Analyze in terms of number of errors detected Deviation method finds 2.5x errors Estimate physical measurement method finds 40% of errors Physical measurement method Deviation analysis method Room # measurements # errors identified # interest regions # errors identified Room 1 13 4 6 17 Room 2 11 5 3 6 TOTAL 24 9 9 23

Summary and Conclusions:

Summary and Conclusions Deviation analysis method has advantages over the physical measurement method: Complete coverage Provides insight into the source, type, and magnitude of errors Deterministic assessment of quality for given accuracy requirements Measures absolute errors rather than relative measurements 23

Future Work:

Future Work 24 Improve existing software – work with vendors Develop automation processes using computer vision algorithms More analysis of registration and calibration errors Extend types of deviation that can be detected (e.g., deviations in the plane of the components) Automatic occlusion map Automatic clutter recognition

Summary of Software Comparison Results:

Summary of Software Comparison Results 25 Evaluation Criteria Deviation Analysis Interoperability Large Data Sets Polyworks Good Good Medium Rapidform Medium Poor Poor Geomagic Poor Good Good Cyclone Poor Poor Good Navisworks Poor Good Good