Doctoral dissertation · UBC

Monitoring and Analyzing Human-Nature Interaction through Sensor and Vision Technologies

A computational, human-centered framework for understanding how people experience urban nature — built in the tradition of those who understood cities by watching the people in them.

Mingze Chen · Faculty of Forestry & Environmental Stewardship, UBC · Supervisor Dr. Keunhyun Park

To understand a place, you watch the people in it. For half a century, urbanists read cities by eye. This dissertation inherits that tradition — and extends it into urban nature, at a scale their stopwatches never could reach.

1960
Kevin Lynch
The Image of the City
People carry mental maps of the city — paths, edges, districts, nodes, landmarks. How a place is read became something you could study.
1961
Jane Jacobs
Death and Life of Great American Cities
Jacobs read the street as an “intricate sidewalk ballet” — a city understood not by its plans, but by how people actually use it.
1971
Jan Gehl
Life Between Buildings
Gehl turned that conviction into method: first life, then space, then buildings — public life as something to be observed and measured.
1980
William H. Whyte
The Social Life of Small Urban Spaces
Whyte sat on rooftops with time-lapse cameras, counting how people really used plazas. Observation became evidence.
2000s →
Computational urbanism
GPS · sensors · vision · AI
Then the instruments changed. Sensors, social media, computer vision and AI let us observe urban life at a scale no notebook could — yet most of it still asks only where people are, not how they engage.
The core
Four dimensions
a way of seeing engagement
This work frames human-nature interaction as a four-dimensional process — moving from counting bodies toward interpreting meaning: , , , and meaning. (tap each)
Chapter 2
Foundation
A scoping review maps how sensor- and vision-based technologies monitor behavior in natural areas — and which dimensions each can see.
Chapter 3
Sensing
A low-cost BLE system measures presence and movement at two field sites — no cameras, validated to ICC .97.
Chapter 4
Vision
A vision-language model learns to read nature as setting, subject, and partner — activity and meaning, not just counts.
Chapter 5
A decade
Applied across a decade of Vancouver's urban nature — 16,276 photographs of how a city actually meets its parks, shores, and trees.
The studies

Three studies, one trajectory

A review maps the terrain; a sensor study measures presence and movement; a vision model reads meaning. Open any to explore its interactive findings.

Chapter 2 · Scoping review

Sensing people in nature

How sensor- and vision-based technologies monitor human behavior in natural areas — and which dimensions each one can actually see.

246
studies coded
~24,530
records screened
3
method families
Explore the review →
20102024PUBLICATIONS BY METHODsensorvis
Chapter 3 · Sensing

Counting a park without watching it

A low-cost Raspberry-Pi BLE system that reliably measures presence and movement at two field sites — without a single camera.

2
field sites
.97
inter-sensor ICC
0
cameras
Enter the signal lab →
S1S2flowBLE · TWO SENSORSpassive device counts → people + movement
Chapters 4-5 · Vision

Reading human-nature interaction

A vision-language model interprets nature as setting, subject, and partner across a decade of Vancouver's urban nature.

16,276
photographs
2014-25
decade
3
interaction roles
See the decade →
NATURE'S ROLE — ONE OF THREEsettingsubjectpartner
“We should look hard, with a clean, clear mind, and then look again — and believe what you see.”
— William H. Whyte
Author

Mingze Chen
Research lead · mingzechen.com

Supervision & committee

Dr. Keunhyun Park — Supervisor
Dr. Guangyu Wang · Dr. Angela Rout
UNDER Lab

Institution

University of British Columbia
Faculty of Forestry & Environmental Stewardship

Mingze Chen · Doctoral dissertation · UBC Forestry & Environmental Stewardship · 2026