Innovative Method for Enhancing Pavement Crack Images

Automated detection of pavement cracks using digital imaging technology is a safe, accurate and speedy alternative to the traditional visual pavement crack evaluation. Once images are collected they must invariably be enhanced using noise filtering techniques to accurately determine the extent and severity of cracks. Sophisticated mathematical filtering techniques use suppression of selected frequencies to filter out noise in images indiscriminately, while simpler statistical filtering techniques are based on statistical modeling of noise. The theoretical assumptions required for modeling noise can affect the accuracy of the latter techniques. Furthermore, in both types of techniques, indiscriminate filtering can result in blurring of the images although adoption of edge detection based on gradient operators can help in retaining some important information of the image. An innovative statistical filtering method based on detailed analysis of noise is introduced to address the above issues and enhance pavement crack images more effectively. The new method involves capturing the image of the Gray-scale target under the same ambient temperature and lighting conditions as those of the crack imaging operation. In this method the extent of noise in a given region of the image is determined using the noise in the image of the Grayscale wedge of the corresponding intensity. Concrete and asphalt pavement crack images enhanced by existing methods based on pixel intensity statistics were compared to those enhanced by the new method. Using subjective visual observation and objective evaluation criteria it was determined that the new method is not only accurate and more effective but also easily implemented in imaging operations.

Language

  • English

Media Info

  • Media Type: CD-ROM
  • Features: Figures; References; Tables;
  • Pagination: 22p
  • Monograph Title: TRB 86th Annual Meeting Compendium of Papers CD-ROM

Subject/Index Terms

Filing Info

  • Accession Number: 01043832
  • Record Type: Publication
  • Report/Paper Numbers: 07-3311
  • Files: TRIS, TRB
  • Created Date: Feb 8 2007 8:00PM