|
NCHRP IDEA 20-30/IDEA 081 [Completed (IDEA)]
Automated Real-Time Pavement Crack Deflection/Classification System
[ NCHRP 20-30 (NCHRP-IDEA) ]
| Project Data |
| Staff Responsibility: |
Dr. Inam Jawed |
|
This project developed an automated real-time crack detection and analysis system based on image processing and computer vision techniques. The system consists of a personal computer, a frame grabber with two on-board processors, a distance sensor and a video camera mounted on top of a van. The images from the video camera are captured and converted to digital images by the frame grabber, while the images are recorded by the video camera for future reference. Over 20,000 images were obtained under different vehicle speeds and light conditions and digitized. Processing algorithms were developed and applied to the collected images. The effectiveness and speed of the algorithms were improved for features such as segmentation, enhancement, noise removal, Hugh transformation and morphology, etc. for crack detection and classification applications. Three evaluation criteria were used: performance for different pavement types including cracks, sealed cracks and shadows, performance under different light conditions and circumstances, and performance when there are some tars (bleeding) or other non-crack scenes on the images. Pavement images were obtained with vehicle speeds of 35 mph to 75 mph under different lighting conditions, including both cloudy and sunny days. The results demonstrate that the proposed system can accurately process the images of different types of pavements and under different lighting conditions, including the shadows (Figures 1 and 2). The project team is working closely with Utah DOT. Several private companies, including ROADWARE and LAW Engineering, have expressed interest in the implementation of this technology. The final report is available from the National Technical Information Service (NTIS # PB2003-101350).
|
|
