Urban Building Usage Labeling by Geometric and Context Analyses of - - PowerPoint PPT Presentation

Institute of Cartography and Geoinformatics | Leibniz Universität Hannover

Hai Huang, Birgit Kieler and Monika Sester

Urban Building Usage Labeling by Geometric and Context Analyses of the Footprint Data

Building Usage Labeling | ICC Dresden, 29th Aug. 2013 Hai Huang | 2

►

The Usage (use and occupancy) information of buildings

• is of great interest for, e.g., navigation, city planning, emergency

management, etc.

• is not available in a consistent way in the volunteered data
• is not always available even in the official cadastral maps

►

Our approach to automatic labeling:

• Pre-defined category: residential (family house, apartment

building), com m ercial (shopping mall, office building), industrial (factory, warehouse) and public (school, hospital, theater, etc.)

• High-level local geom etric features and contextual constraints
• Markov Random Field (MRF) for the modeling of the building

netw ork

Introduction

Building Usage Labeling | ICC Dresden, 29th Aug. 2013 Hai Huang | 3

►

Influence of the local geometric features

• Footprint area
• (+ ) distinguish e.g., industrial and single-family house
• (-) hard to separate public and industrial building without

considering the shape

• Length/ width ratio
• (+ ) distinguish e.g., bar shape (often industrial, residential)

and square shape (often public and commercial)

• (-) only works for simple shapes (or using bounding box),

cannot represent complex buildings  Higher level features are required to integrate multiple geometric attributes and provide more precise description to the footprint shape.

Local features

Building Usage Labeling | ICC Dresden, 29th Aug. 2013 Hai Huang | 4

►

Effective Width (EW)

• Definition: the average width of the footprint along the centerline
• f the footprint

AB

: building area

LB

: actual building length : length of the modified center line (red)

Local features

~

Building Usage Labeling | ICC Dresden, 29th Aug. 2013 Hai Huang | 5

►

Effective Width (EW)

• Meaning: the general living/ movement space inside the building!
• …

despite the building complexity

Local features

Building Usage Labeling | ICC Dresden, 29th Aug. 2013 Hai Huang | 6

►

Branching degree (BD)

• Definition: a score of the number and distribution of the building

segments (the longest: “trunk”; the others: “branches”)

• Basis: conventional straight skeleton

Local features

Building Usage Labeling | ICC Dresden, 29th Aug. 2013 Hai Huang | 7

►

Branching degree (BD)

• Meaning: a measurement of the building complexity

Local features

Building Usage Labeling | ICC Dresden, 29th Aug. 2013 Hai Huang | 8

►

The local energy

• EW-BD distribution
• Each building can be represented as a point in this parameter

space.

• The class centroids are empirically given with generic values.

Local features

Building Usage Labeling | ICC Dresden, 29th Aug. 2013 Hai Huang | 9

►

The local energy

• The probability that this building belongs to one of the classes is

inversely proportional to its (standardized) distance to the centroid of the class.

• Energy: a quaternary value of the probabilities (sum= 1)

I.e., the probabilities that this building is supposed to be labeled to all the individual classes.

• E.g.,

implies the building will be more likely labeled as “industrial” considering only the local features.

Local features

Building Usage Labeling | ICC Dresden, 29th Aug. 2013 Hai Huang | 10

►

The neighborhood relationship

• Voronoi cells as basis

Context model

Building Usage Labeling | ICC Dresden, 29th Aug. 2013 Hai Huang | 11

►

Markov Random Field (MRF)

• Vertices: centroids of individual buildings
• Edges: neighborhood relationship

Context model

Building Usage Labeling | ICC Dresden, 29th Aug. 2013 Hai Huang | 12

►

The total energy

• Unary term
• Binary term

Context model

N R C I P R 1

• 1

0.5 C . 1 0.5 I . . 1

• 1

P . . . 1

Building Usage Labeling | ICC Dresden, 29th Aug. 2013 Hai Huang | 13

►

A global optimization

• The configuration K, which leads to the maximum total energy
• Four possible labels for each building/ vertices
• Change of each label leads to a new configuration
• Urban area 

highly connected network  Computational intractable for direct solution…

Context model

Building Usage Labeling | ICC Dresden, 29th Aug. 2013 Hai Huang | 14

Stochastic search

Initialization M0 Propose a new state M’ Sample lables for the buildings Calculate the overall energy Accept the new proposal (Metropolis-Hastings criterion) Mn+ 1= M’ Mn+ 1= Mn Y N current Mn

Building Usage Labeling | ICC Dresden, 29th Aug. 2013 Hai Huang | 15

►

Dataset 1 (OSM): 94 buildings, Boston, United States

Experiments

Building Usage Labeling | ICC Dresden, 29th Aug. 2013 Hai Huang | 16

►

Ground truth

Experiments

Residential Commercial Industrial Public

Building Usage Labeling | ICC Dresden, 29th Aug. 2013 Hai Huang | 17

►

Labeling using only local features (Accuracy= 72.3% )

Experiments

Residential Commercial Industrial Public Incorrect classifications

Building Usage Labeling | ICC Dresden, 29th Aug. 2013 Hai Huang | 18

►

Final results using both local and contextual information (Accuracy= 97.8% )

Experiments

Residential Commercial Industrial Public Incorrect classifications

Building Usage Labeling | ICC Dresden, 29th Aug. 2013 Hai Huang | 19

►

Dataset 2 (cadastral map): 456 buildings, Hannover, Germany (Accuracy= 89.7% )

Experiments

Residential Commercial Industrial Public Labeling result Ground truth Incorrect classifications

Building Usage Labeling | ICC Dresden, 29th Aug. 2013 Hai Huang | 20

►

A novel full automatic labeling of building types (use and occupancy)

►

Solely based on the building footprint data

►

Four-classes category: residential, commercial, industrial and public

►

Two new high-level geometric local features: effective width and branching degree

►

MRF for the modeling of building network, integrating contextual constraints

Conclusion

Building Usage Labeling | ICC Dresden, 29th Aug. 2013 Hai Huang | 21

►

A general, rather rough category

►

… against the wide variety of building types and definitions

►

Some building types, e.g., educational, high-hazard, can actually NOT be derived solely from footprint data.

►

Finer contextual knowledge can be integrated. E.g., an identified market square may increase the likelihood that the surrounding buildings are commercial!  New local features as well as contextual knowledge can be easily added in the proposed framework… for a refined or specialized classification.

►

A general classifier is used all the datasets. In the future work, the local/ specified parameters of the classifier can be learned for individual cities.  A “spectral signature” of the city…

Problems & outlook

Building Usage Labeling | ICC Dresden, 29th Aug. 2013 Hai Huang | 22

Thank you very much for your attention!