Student research opportunities

Image-Based Car Damage Detection (ICAR)

Project Code: CECS_666

This project is available at the following levels:
Engn4200, Honours, Masters

Keywords:

computer vision; image processing; pose estimation; 2D-3D registration; segmentation; interactive tool.

Supervisors:

Dr Stephen Gould
Professor Marcus Hutter

Outline:

Today experts judge car damages on-site and from photos. This is expensive, slow and subjective. A (semi)automatic estimation of the repair-costs is desirable. Under idealized conditions of normalized full-scale car images, a mapping and comparison of broken to unbroken cars would allow an automatic estimation of car body damages. Under real conditions, this is a very challenging task. The project aims at developing a (semi)automatic support for damage evaluation. It involves various hard computer vision [FP02] and image processing [GW07] problems: Full pose estimation by automatically registering a 3D model to high accuracy onto a single 2D photo from an uncalibrated camera in an unknown location and lighting condition [HB09,JHB10], segmenting a car into its panels, detecting damage on shiny panels that mirror the environment, and others.

Goals of this project

• integrate existing algorithms into one interactive tool
  
• develop and add (some smaller/feasible) additional functionality
  
• do extensive experimentation and testing and interpretation of results
  
• write a tutorial and a user manual
  

Requirements/Prerequisites

• excellent software development skills, in C(++)
  
• ideally experience with computer vision libraries and GUI toolkits
  
• good writing skills
  

Student Gain

• getting acquainted with modern computer vision algorithms
  
• handling a complete project from specification to implementation to testing to documentation
  

Background Literature

• [FP02] D. A. Forsyth and J. Ponce.
  Computer Vision: A Modern Approach.
  Prentice Hall, 2002.
  
  
• [GW07] R. C. Gonzalez and R. E. Woods.
  Digital Image Processing.
  Prentice Hall, 3rd edition, 2007.
  
  
• [HB09] M. Hutter and N. Brewer.
  Matching 2-d ellipses to 3-d circles with application to vehicle pose estimation.
  In Proc. 24th Conf. on Image and Vision Computing New Zealand (IVCNZ'09),
  pages 153--158, Wellington, New Zealand, 2009. IEEE Xplore.
  

Contact: