Buildings3 d gThe Use of Geographical Information Systems for 3D Urban Models Reconstruction from Aerial Lidar Data

21/4/2015 Dr. Ahmad Yousef

The Use of GIS for 3D Urban ModelsReconstruction Based on Aerial Lidar Data

Dr. Ahmad Yousef


• Overview

• Objectives & Goals

• Digital Terrain Model Extraction

• 3D Building Model Reconstruction

• Conclusion & Final Thoughts

AGENDA


Pollution, CairoVisualization, Dubai

Pollution, Moscow

Flood, Pakistan

Visualization, Los Angeles

City Planning, Stuttgart

Flood, Mexico

• GIS has traditionally been 2D technology.

• New product and technology make us reconsider the role of 3D.

• It is widely recognized that 3D models are necessary.

OverviewWhy 3D GIS …


Year : 2050

66%

Year : 2014

54%

Urban Population

World

Source: UN DESA’s Population Division

of Global Population

o The overall growth of theworld’s population could addanother 2.5 billion people tourban populations by 2050

o 90 percent of the increaseconcentrated in Asia andAfrica

OverviewWhy urban area …


OverviewWhy lidar data …

Waldkirch, Germany


OverviewWorkshop Objectives & Goals





1. Introduce a fast and simple integrated Digital Terrain Model (DTM) extraction framework in a GIS environment.



1. Introduce a fast and simple integrated Digital Terrain Model (DTM) extraction framework in a GIS environment.

2. Introduce a developed GIS approach for reconstructing of the 3D building models from lidar point clouds.


Digital Terrain Model Extraction

How to classify lidar data into terrain and off-terrain points?


• Filtering means classification of points into terrain and off-terrain.

• Bare earth is assumed to be continuous surface.

• Filtering In/Out data:

– Point list

– Grid

– Triangulated Irregular Network - TIN

DTM ExtractionFiltering lidar data


DTM ExtractionEsri Terrain Dataset

Surface geometry at different resolution (TINs)

– TIN is acronym for triangulated irregular network.

Level of Detail-Pyramid

LoD x

Pyramid Type:I. WINDOWSIZEII. ZTOLERANCE

parameter:I. ZMINII. ZMAXIII. ZMEANIV. ZMINMAX










DTM ExtractionMinimum Block Classification - MBC

Window Block

Bare Earth Surface

Input Points Dataset [Geodatabase], Input Window Size, Cell Size (≈ Points Average Distance)

Output Ground Points, DTM

Window Block





1

2

3

5

Window Size Z-threshold

32 3.2

16 1.6

8 0.8

4 0.4

2 0.2

4





Convert Terrain to TIN

Smoothing

Umbrella Filter


DTM ExtractionDigital Surface Model


DTM ExtractionDigital Terrain Model


Digital Surface Model Shaded Relief

DTM ExtractionForest Dataset …

Digital Terrain Model Shaded Relief


45m x 35m

DTM ExtractionUrban Dataset …

5 750 000 Points


DTM ExtractionUrban Dataset …


3D Building Models Reconstruction

How to fulfill the gap between lidar data and building models ?How to minimize the number of points that represent a building ?


3D Building ModelBuildings Shapes …

Redlands, USA. Source : Google maps


3D Building ModelNo Comments …

Venezuela. Barrio Petare, Caracas.Source: A sustainable approach to problems in urban squatter developments


3D Building ModelLevel of Details LoD

• Building model is the representation used for describing the form of building.

• The complexity of a 3D building model is known as the level of details (LoDs).

Source: : Open Geospatial Consortium CityGML Implementation Specification 1.0,20.8 2008

Level of Details

Data Processing

LoD 1 : Flat Roofs

LoD 2 : Roof Type

LoD 3 : Real Roof Shape

LoD 4 : Interior


Model Structure Based – Model Derive

3D Building ModelModeling Approaches

• Model Database.

• The final roof shape is always topologically correct.

• Complex roof shapes cannot be reconstructed.

3D Building Models Database

Flat Desk Gable Hipped Mansard PyramidGambrel



Source: International Summer School “Digital Recording and 3D Modeling”.

3D Building Models Database

Flat Desk Gable Hipped Mansard PyramidGambrel


Data Structure Based – Data Derive


• Roof described by planar faces.

• Partitioning the given ground plan and find the most appropriate plane segment to each partition.

Source: International Summer School “Digital Recording and 3D Modeling”.


Input for Data Derive Model


I. Points based

– Points may belong to several planes.

II. Raster based

– Information content is decreased due to interpolation.

III. TIN based

– To avoid loss of information due to interpolation, all operations are performed on the Delaunay triangulation of the original height points.

– Requires more analysis.


3D Building ModelBuilding Model Elements

Planar Patches

Wall

Footprint

Planar Face

TIN Triangles

Lidar

Roof Boundary


5 Processing Steps :-

3D Building ModelFramework …

1. Generate Triangle Irregular Network - TIN

2. Extract Roof Planar Patches

i. Normal Vector Estimation

ii. Segmentation & Region Growing

3. Detection of Planar Roof Faces

i. Least Square Plane Fitting

ii. Merging Planar Patches

4. Intersection of Roof Planes

5. 3D Model Reconstruction



3D Building Model1 - Triangle Irregular Network


3D Building Model1 - Triangle Irregular Network


3D Building Model2 - Extract Roof Planar Patches

A.X + B.Y + C.Z + D = 0N=(A,B,C)



Normal Vector Estimation

1 Ring Neighborhood 2 Rings Neighborhood








1 Ring Neighborhood






2 Rings Neighborhood


2 Rings Neighborhood





Segmentation & Region Growing

Apply region growing to find roof patches











3D Building Model3 - Detection of Planar Roof Faces



Vertical Triangles

All triangles with slope greater than 60 are considered as vertical patches



Least Square Plane Fitting

A * X + B * Y + C * Z = D

The Plane Normal is given by : N = (A,B,C)



Intersection of adjacent patches

A * X + B * Y + C * Z = D



Merging Planar Patches


3D Building Model4 - Intersection of Roof Planes

Patch 1 2 3 4 5 6 7

1 YES YES YES YES

2 YES YES YES

3 YES YES

4 YES YES

5 YES YES

6 YES YES YES

7 YES

Plane Adjacency Matrix






3D Building Model5 - 3D Model Reconstruction












3D Building ModelExample 1







Minimum area threshold = 5 m2

Minimum area threshold = 3 m2




3D Building ModelTest Area


3D Building ModelError Sources


SummaryConclusion & Final Thoughts...

• Minimum Block Classification (MBC) Model was successfully implemented in GIS environment.

• Advantages of MBC Model includes :

– Fixed number of processing loops.

– Capable of capturing and removing the major terrain features.

– Although building size and shape present a challenge for many other filtering

algorithms, they do not significantly hinder the MBC algorithm when using the

proper window size and threshold.


SummaryConclusion & Final Thoughts...

o Minimum Block Classification (MBC) Model was successfully implementedin GIS environment.

o 3D building reconstruction models from lidar data was developed fromconstructing a roof surface geometry.

o 3D building reconstruction models result affected by

Minimum area threshold

Points distribution

Shape complexity

o The processing time varies and is dependent on the shape of the building,start from 5 seconds for simple buildings to 30 seconds for complexbuildings.


Thank YouShare What You Can To Benefit The OthersAhmad Yousef ([email protected])

Buildings3 d gThe Use of Geographical Information Systems for 3D Urban Models Reconstruction from Aerial Lidar Data

Engineering

ahmad yousef filtering

ahmad yousef pollution

ahmad yousef year

terrain points

dtm window block

classification of points

input window size

d models