11-668 - Adaptive Vision-Based Crack Detection with Depth Perception

Description:
  • Civil infrastructure inspection
  • Building inspection

Abstract

USC researchers introduce a vision-based crack detection methodology that processes 2D digital images by analyzing the scene's geometry. The methodology incorporates feature extraction and classification techniques to distinguish cracks from non-crack patterns. Experimental tests have been conducted to evaluate the system's performance, including validation tests with real concrete cracks. The system's ability to extract the entire crack makes it suitable for integration with autonomous or semi-autonomous robotic systems, enabling crack thickness quantification.

Benefit

  • Automated crack detection
  • Integration with robotic systems
  • Improved accuracy, cost, and time
  • Crack thickness quantification

Market Application

Current standards for structure condition assessment require labor- intensive and subjective visual evaluations by inspectors. Civil infrastructure systems represent a significant portion of global assets, estimated at $20 trillion in the United States alone. With over 10,400 structurally deficient bridges, there is an urgent need for effective inspection and evaluation approaches. A robotic system to autonomously inspect structures and detect damage could offer a less time-consuming and expensive solution while eliminating human error and prolonging the service life of infrastructure assets.

Publications

An innovative methodology for detection and quantification of cracks through incorporation of depth perception, Jahanshahi et al., 2013.

Other

Stage of Development

  • Tested with real concrete cracks
  • Available for licensing

Patent Information:

  • Title: Image Based Crack Quantification
  • App Type: Utility
  • Country: United States
  • Serial No.: 13/567,969
  • Patent No.: 9,235,902
  • File Date: 8/6/2012
  • Issued Date: 1/12/2016
  • Expire Date: 5/9/2034
  • Patent Status: Patent Issued