Aerolyzer

Name: Kim Whitehall
Affiliation: NASA JPL
Phone: 3217045011
E-mail: kwhitehall@gmail.com
Website: https://aerolyzer.github.io/
Knowledge Required: An interest in image processing and analysis technologies. Candidates should be fluent in Python and proficient in using *nix environments. Candidates should be willing to learn data mining and data fusion techniques as well as atmospheric optics and basic meteorology. Candidates should also some prepared to learn about and participate in open-source communities. Note that all code will be permissively licensed under the Apache Software Foundation v2.0 license.
Motivation: For the average citizen, getting accurate real-time updates of atmospheric conditions is a challenge. This difficulty results in limited availability of information about current aerosol content due to limited distribution of instrumentation. Meanwhile, millions of digital images are taken on mobile phones daily, some of which contain outdoor scenes. Combining this volume of mobile users’ landscape images with camera metadata and existing atmospheric information can provide an untapped resource for inferring atmospheric phenomenon. The purpose of this project is to develop an application that supports the ability to understand content from these digital outdoor images in order to better monitor current aerosol content. This project is a continuation of the Aerolyzer Capstone 2016-2017 project that will deliver the aerolyzer Python library by improving on the existing library to include algorithms to identify atmospheric phenomena and characterize atmospheric composition.
Description: The ability to get real-time updates of atmospheric air quality conditions can be very limited. There are few instruments (ground-base, in-situ or satellite-based) that make these recordings. However, it is proposed that the non-conventional source of photographs can be leveraged to infer basic knowledge of the air-quality given the principles in atmospheric optics and the governing equations of transport. With many emergency response organizations and media houses recording images of a field of view consistently throughout the day, a method to identify (weather) phenomena associated with aerosol changes e.g. smoke stacks, within images would be useful (to trigger action/response). Furthermore, the general public could benefit through an air quality application that leverages photos they take and contribute. Currently, the available image feature detection software uses a standard shape definition to identify features. While this will be useful, there is a need for a tool that enables feature detection based on an image’s color distribution in order to infer the atmospheric phenomena and composition of atmospheric aerosols. Images and other data from available sources will also be leveraged to assist with inferring atmospheric aerosol composition, as well as aid in determining the type of atmospheric phenomenon the image displays (for example, a sunset or sunrise).
Objectives: (1) Improve on the existing Aerolyzer application (http://ec2-54-152-58-132.compute-1.amazonaws.com/app/) through the development of the aerolyzer Python library. (2) Incorporate other data sources to assist with the interpretation of the images regarding aerosol content. (3) Provide a set of visualizations and metrics for reporting the findings of the aerosol analysis. (4) Provide examples of how to use the library in a web-based notebook environment (specifically Apache Zeppelin).
Deliverables: The purpose of this project is to develop a library capable of processing visible images and inferring atmospheric (optical or otherwise) phenomena. The core deliverable is the development of the Python aerolyzer library to perform the analysis. The secondary deliverable is to integrate the library into a web-based application.
Other comments: This project will be co-mentored with Dr. Lewis McGibbney.
Very research oriented! Come with an open mind, and we can learn together what is needed to build this application.

   D. Kevin McGrath
   Last modified: Thu Nov 16 11:32:03 2017