University of Illinois Urbana-Champaign

Quantifying the impact of fog on autonomous driving object detectors and developing a fog-aware vehicle detector

Kore, Ruhi

This item is only available for download by members of the University of Illinois community. Students, faculty, and staff at the U of I may log in with your NetID and password to view the item. If you are trying to access an Illinois-restricted dissertation or thesis, you can request a copy through your library's Inter-Library Loan office or purchase a copy directly from ProQuest.

Permalink

https://hdl.handle.net/2142/120450

Description

Title
Quantifying the impact of fog on autonomous driving object detectors and developing a fog-aware vehicle detector
Author(s)
Kore, Ruhi
Issue Date
2023-05-04
Director of Research (if dissertation) or Advisor (if thesis)
Forsyth, David
Department of Study
Computer Science
Discipline
Computer Science
Degree Granting Institution
University of Illinois at Urbana-Champaign
Degree Name
M.S.
Degree Level
Thesis
Keyword(s)
  • fog
  • object
  • vehicle
  • detector
  • detectron
  • renderings
Abstract
An essential component of developing robust autonomous driving software is the ability to successfully navigate in various weather conditions, such as in fog, rain, and snow. Autonomous vehicles today rely on hardware (e.g., cameras) and sensors (e.g., lidar, radar) to understand their environment so they can navigate their surroundings accordingly. However, severe weather impacts the quality of data obtained by the hardware and sensors. For instance, in foggy weather conditions, the contrast in the images obtained by cameras drops significantly, making it difficult for intelligent image processing algorithms to perform object detection and image classification. In this project, we quantify the impact of fog on the accuracy and confidence levels of current state-of-the-art object detectors, focusing on the task of identifying other vehicles on the road in foggy weather conditions. We do so by curating a dataset of road images from the driver’s perspective, with various levels of fog synthetically added to each image. We also design a vehicle detector that can identify vehicles in fog with a higher accuracy and confidence level compared to current state-of-the-art detectors. Our fine-tuned detector’s persistence in correctly identifying vehicles is, on average, 4.69% higher in light fog, 13.38% higher in medium-intensity fog, and 23.65% higher in heavy fog.
Graduation Semester
2023-05
Type of Resource
Thesis
Copyright and License Information
Copyright 2023 Ruhi Kore

Owning Collections

Graduate Dissertations and Theses at Illinois PRIMARY
Graduate Theses and Dissertations at Illinois