Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
AI Analysis of City Changes
同济大学智能规划协同创新中心
China Intelligent Urbanization Co-creation Center
iCity Group
奥泰因.赫尔佐格Prof. Dr. Otthein HERZOG
德国国家工程院院士Academician, acatech
United Nations World Geospatial Information Congress
Deqing, November 19, 2018
1. City Development Challenges and Artificial Intelligence
2. Analyzing and Classifying City Images with Deep Learning Methods(Work done at CIUC by iCity Group)
3. Examples
Outline
Challenges for Intelligent Cities: Activity Spheres
Energy and resource
infrastructure
Service and
attendance
Politics and
administration
Health
Education
Communication/IT
Human and
lifestyle
Urban Production
Nutrition
Environment
and climate
Transport and Logistics
Buildings and
living
Mobility
and traffic
City structure
Convergence of city
systems
Safety
AI system capabilities are:
1.Reasoning, Problem Solving, Decision support
2.Knowledge Representation/Engineering
3.Learning/Big Data Analytics
4.Planning
5.Natural Language Processing
6.Perception – Pattern Recognition – Multimodal Interfaces
7.Motion and Manipulation/Robotics
8.Social Intelligence
9.Creativity
Artificial Intelligence Areas
Source: Wikipedia, “Artificial Intelligence”
The overall research goal of Artificial Intelligence is to create technology that allows computers and machines to function in an intelligent way.
AI system capabilities for Urban Planning are:
1.Reasoning, Problem Solving, Decision support
2.Knowledge Representation/Engineering
3.Learning/Big Data Analytics
4.Planning
5.Natural Language Processing
6.Perception – Pattern Recognition – Multimodal Interfaces
7.Motion and Manipulation/Robotics
8.Social Intelligence
9.Creativity
Artificial Intelligence Areas
Source: Wikipedia, “Artificial Intelligence”
The overall research goal of Artificial Intelligence is to create technology that allows computers and machines to function in an intelligent way.
7
“Traditional” Software Systems
1 Sense• Audio (Speech)• Video (Photo, Movie)• Digital (Log Data)• IoT (Sensor Data)• Touch (Fingerprint)• Smell (Gas Detection)
2 Comprehend• Specific interpretations of
input data
3 Act• Command Appliances• Trigger follow-up processes• Display meaningful texts• Play translated speech• Draw printings• Compose music
After Purdy, M., & Daugherty, P. (2016). Why Artificial Intelligence is the Future of Growth. Abgerufen am 8. Juni 2017 von https://www.accenture.com/lv-en/_acnmedia/PDF-33/Accenture Why-AI-is-the-Future-of-Growth.pdf
8
Knowledge-based Artificial Intelligence Systems
1 Sense• Audio (Speech)• Video (Photo, Movie)• Digital (Log Data)• IoT (Sensor Data)• Touch (Fingerprint)• Smell (Gas Detection)
Explicit Application Domain Knowledge acquired through• Deep Learning• Machine Learning• Statistics/Analytics• Rule Sets• Ontology development
2 Comprehend• Deep Learning• Machine Learning• Statistics/Analytics• Rule Sets• Decision Making
3 Act• Command Appliances• Trigger follow-up processes• Display meaningful texts• Play translated speech• Draw printings• Compose music
After Purdy, M., & Daugherty, P. (2016). Why Artificial Intelligence is the Future of Growth. Abgerufen am 8. Juni 2017 von https://www.accenture.com/lv-en/_acnmedia/PDF-33/Accenture Why-AI-is-the-Future-of-Growth.pdf
9
Learning Artificial Intelligence Systems
1 Sense• Audio (Speech)• Video (Photo, Movie)• Digital (Log Data)• IoT (Sensor Data)• Touch (Fingerprint)• Smell (Gas Detection)
5 Application Domain Knowledge: Learn /Train• From coding to training
2 Comprehend• Deep Learning• Machine Learning• Statistics/Analytics• Rule Sets• Decision Making
3 Act• Command Appliances• Trigger follow-up processes• Display meaningful texts• Play translated speech• Draw printings• Compose music
4 Feedback• Cause/effect• Right/wrong
After Purdy, M., & Daugherty, P. (2016). Why Artificial Intelligence is the Future of Growth. Abgerufen am 8. Juni 2017 von https://www.accenture.com/lv-en/_acnmedia/PDF-33/Accenture Why-AI-is-the-Future-of-Growth.pdf
Technological Innovation:
World Largest City Big Data Base constantly updated技术创新:建立世界城市大数据,并持续升级 追踪超越原最大数据库
1988 1997 2016
4 Generations of City Big
Data Base
Quantitative Analysis
of World Cities
2007
Media-Scale Cities
Simulation• 12 Billion data records in 8 Categories• 集聚了8类120亿条有效数据
美国UIC城市诊断
数据库
团队建构的WUBDB
数据库
120亿
有效
数据
2亿
有效
数据
1. City Development Challenges and Artificial Intelligence
2. Analyzing and Classifying City Images with Deep Learning Methods(Work done at CIUC by iCity Group)
3. Examples
Outline
City Tree“城市树”
Satellite Images of Cities through 40 years跨度40年城市卫片
Super Speed: Intelligent Identification速度超越:智能识别
Super Precision:30mX30m精度超越:30X30m
EU 1X1km2
欧盟 1X1km2
New representation: City Tree首创“城市树”
2015
2005
2000
1995
1990
1985
1980
1975
S W E N
2010
Spatial Law空间规律
Ningbo
Spatial Law空间规律
纽约
北京
Spatial Law空间规律
937个美国 200个
2036个50km2以上的城市
100km2以上的城市
3520个30km2以上的城市
5404个20km2以上的城市
9043个10km2以上的城市
9516Cities over 9km2
10002Cities over 8km2
11071Cities over 5km2
11663Cities over 4km2
13343Cities over 2km2
通过机器识别技术超越美国最大城市数据库原有的200个城市人工识别记录
Cities over1km2 13810
City Trees for 13810 world cities have been diagrammed by 2018 Jan 18至2018年已绘制13810个城市的城市树
全样本 全球范围内识别独立建成区13810个
Full coverage of 13810 independent built areas in the world
year
Area km2
2
2 2 2
1
2y a x b x c= + +
2
1 1 1
1
2y a x b x c= + +
Area2015
Area1975
Classifying Cities by Urban Edge Growth按城市边界增长分类城市的方法
Sprouting萌芽型
Rickets佝偻型
Steady Growth成长型
Matured成熟型
Rapid Growth发育型
Regional区域型
Declining衰退型
Illustrating the Development courses of 13810 world’s cities by City Trees绘制“城市树” 完成了全球13810个城市发展历程的识别
Findings: 7 Categories of Urban Development Modes and the related laws of growth. 发现并定义为7类城市发展类型及其特性,以及城市成长规律。
China
Brazil
Argentina
UK
Germany
France
Matured成熟型
1900
Declining衰退型
201
Spouting萌芽型
435
Regional区域型
143
Rickets佝偻型
3601
Steady 成长型
2365
Rapid 发育型
831
7 Categories of cities by the measure of “City Tree”
7种类型的城市分布
Yangtze River Delta region
1975
1990
2000
2015
◼ 苏州、上海、无锡、常州、南通城镇群集聚历程Urban Agglomeration Process
1. City Development Challenges and Artificial Intelligence
2. Analyzing and Classifying City Images with Deep Learning Methods(Work done at CIUC by iCity Group)
3. Examples
Outline
长三角城市群 京津冀城市群Beijing, Tianjin and Hebei city clusterYangtze River Delta city cluster
长江中游城市群 粤港澳湾区Middle Yangtze River City Cluster Guangdong, Hong Kong and Macao Bay Area
美国西海岸 美国东海岸U.S. West Coast U.S. East Coast
东京湾五大湖U.S. The Great Lakes region Tokyo Bay Area