Heinz Platform Pittsburgh:
Platform Pittsburgh: Air Quality Data Extraction from Images

Srinivasa Narasimhan, Illah Nourbaksh

:: Challenge :: Air Quality Monitoring :: Camera as an Air Quality Sensor ::
:: Take Action :: News :: Acknowledgements :: Contact ::

Several Examples of Air Quality Captured in Digital Imagery


Air pollution is a growing problem negatively impacting the health and lives of people worldwide. Over 3 million deaths in the world are due to air pollution and 25% of emergency room visits are directly related to air pollution. Moreover, asthma is the most chronic disease in children.

Fine inhalable particles with a diameter less than 2.5 micrometers (PM2.5) come from a variety of sources such as power plants, motor vehicles, airplanes, residential wood burning, forest fires, agricultural burning, volcanic eruptions, and dust storms. Some PM2.5 particles are emitted directly into the air while others are formed through interactions in the atmosphere, which is the main cause of haze in the United States.

No levels of PM2.5 particles are truly safe. Because the particles are so small and light 1) they stay in the atmosphere longer increasing the chances of inhalation by humans and 2) may penetrate deep into the lungs and get into the bloodstream. Numerous studies have found a close link between PM2.5 particles and premature death from heart and lung disease, asthma, heart disease, chronic bronchitis, emphysema and pneumonia.

Despite the major health risks associated with poor air quality, there are only a thousand air quality sensors across the United States. However, there are approximately 30 million surveillance cameras and 200 million people with smart phones in the U.S. We propose the development of computational methods for measuring air quality from digital images, a system for air quality monitoring, and a distributed vision-based test bed.

With the ubiquity of cameras in everyday life, our work has the potential to give people air quality measurements anytime, anywhere.

Air Quality Monitoring System

monitoring1_th.png monitoring2_th.png Overview: A community in Neville Township was concerned about the emissions coming from a factory used to refine coal into coak (a fuel for making steel). The community and coke refinery sit in a valley at a lower elevation and surrounded by high elevation areas. Sensors and cameras are deployed in locations volunteered by community members.

The entire system uses an iterative design process. An air quality monitoring system, integrates images from live cameras, particulate matter readings from sensors, and crowdsourced smell reports.

"Smell Reports" are collected from the community about abnormal smells via an online form. In return, we provide scientific evidence and a platform with rich visualizations for citizens to explore data.
Results: Below are examples of images captured at one of our sites. Data are continuously recorded. Previously, community participants spent a lot of time searching for images with emissions. With the collection of more data at more sites, it was becoming a challenge to keep the community engaged in participation. Computer vision methods were developed to detect emissions and significantly reduce time requirements.

aq_image1_th.png aq_image2_th.png aq_image3_th.png

Our air quality monitoring system (video below) provides a time lapse viewer, calendar, sensor and smell reports, smoke detection, and reporting functionality for the user. Regulators from the EPA and ACHD were shown results and that emissions affected the local air quality. Regulators publicly responded to the problem.

The acting director of the EPA from the Region III Air Protection Division in Philadelphia pointed at the screen and said: "But what I see in the video, is totally unacceptable." The local air quality problem became available for further debate and investigation. After the community fought for decades for improved air quality, the system was launched.

Systems are also deployed in Squirrel Hill, North Braddock, and Clairton. 24-hour monitoring also occurs at the Clairton Plant coke works, Braddock's Edgar Thomson Plant, and West Mifflin's Irvin Plant.

back to top

Cameras as an Air Quality Sensor

Tomographic methods permit air quality assessment from digital imagery. Assessing pollution in industrial environments is our current application case. We have tested our system at Haifa Port in Israel.

industrial_pollution aq_sensor.png
sensor_exp.png turbulence.png
Methods were also developed to automatically detect visual air quality cues from live video feeds. Examples below of dust and smoke in construction zones.

 back to top

Take Action

The Breathe Cams are high-resolution, zoomable, 24-hour live camera feeds of Pittsburgh's skyline, the Mon Valley and the Ohio River Valley to help you discover more about the air you breathe. Explore the Breathe Cams, take the tutorial and learn how to capture your own video footage and take action to protect our most precious natural resource - air!

Smell Pittsburgh is a related project that crowdsources smell reports to track how pollutants travel through the Pittsburgh air. You can visualize and download smell reports and learn more at the project website. You can even download the smartphone application to submit your own smell reports!
back to top

In the News

back to top


This research was funded in part by The Heinz Endowments.
back to top