Characterization of Texture Properties of Pavement Images as Aid to Automated Comprehensive Pavement Evaluation

Saumya Amarasiri, Manjriker Gunaratne, Sudeep Sarkar, Abdenour NazefTransportation 2008

Abstract

Automated analysis of pavement images collected by survey vehicles equipped with digital cameras is increasingly becoming popular in pavement evaluation. The majority of the existing analytical techniques used in pavement digital image processing are focused on detecting pavement cracks through automation. In these analytical techniques, it is usually assumed that cracks have darker pixel intensities compared to the background of the image in the gray-scale of intensities from 0 to 255. However, since the pavement surfaces also contain other dark features it has become quite a challenge to differentiate cracks from other extraneous features of the pavement surface. Hence the authors feel that it would be beneficial to characterize the optical texture of pavement images to identify such extraneous features prior to attempting to detect cracks. Tools for characterizing optical surface texture of pavement images have at least three other significant applications; (1) they can be applied to automatically evaluate pavement images with respect to other distress types like rutting, faulting, spalling and loss of skid-resistance, (2) they can be used to monitor pavement deterioration with respect to traffic and environmental conditions, and (3) they can be used as indicators of the quality of imaging conditions with respect to adequate pavement illumination etc. This paper focuses on identifying definitive analytical tools for characterizing the optical texture properties of digital concrete pavement images. The particular images used in this study were collected by the Florida Department of Transportation’s Multi Purpose Survey Vehicle. Digital image processing techniques such as binarizing and moments, quantitative texture measures such as autocorrelation functions and Laws' texture measures, image filtering techniques such as Fourier analysis and power spectral density, as well as the statistically based principal component analysis have been used in this work to characterize optical surface texture of pavement images. Results obtained from most of the above analytical techniques enabled effective characterization of specific texture properties of concrete pavement images. Based on these preliminary findings and with the anticipated extension of the above techniques to asphalt pavements as well, the authors envision the development of a general set of pavement texture characterizing tools that would not only improve the reliability of automated pavement distress surveys but also expand its scope to include the multiple distress types.
BibTeX
@inproceedings{amarasiri2008characterization,
  author = {Saumya Amarasiri and Manjriker Gunaratne and Sudeep Sarkar and Abdenour Nazef},
  title = {{Characterization of Texture Properties of Pavement Images as Aid to Automated Comprehensive Pavement Evaluation}},
  booktitle = {Transportation},
  year = {2008},
}

Loading related content...

Figure