Development of a new digital pavement image processing algorithm for unified crack index computation

In the past, subjective ratings of pavement distress by manual survey methods were used by most public agencies resulting in inconsistent, inaccurate, and unreliable pavement distress data. With an automated digital image processing system for crack analysis, more areas can be surveyed and the collected data can be analyzed quickly. The system can improve the objectivity, accuracy, and consistency of crack survey data. Furthermore, an automated system can reduce the risk of traffic hazards to field inspectors.; The main objective of this research is to develop a new digital image processing algorithm to compute a unified crack index. The image processing algorithm includes the image acquisition, image enhancement, image segmentation, and unified crack index computation. Pavement surface images covering an area of 2.13 m by 1.52 m were collected by a digital camera mounted on a van. The digital images were taken at 30-m stations and stored onto memory cards. The color images were then transferred to a computer, and were converted to ones with 256 gray scales. The images were then archived on a 650 MB compact disc, where approximately 1,500 images can be stored.; Each image with 640 x 480 pixels was divided into 140 smaller tiles where each tile consisted of 40 x 40 pixels. Median filtering method was adopted to suppress noise and enhance the pavement images, while a variable thresholding segmentation technique was used to distinguish cracks from background noise. A minimum number of black pixels was determined to classify a 40 x 40 pixel tile as a crack tile or not. The program then automatically counts the number of crack tiles and computes a unified crack index (CI) for each pavement image. The CI is defined as the ratio of the number of crack tiles to the total number of tiles in a pavement image, expressed as a percentage.; The CIs computed from the Automated Pavement Imaging (API) system were compared with those computed from an independent manual rating procedure used by the Utah Department of Transportation. Good agreement was found between the results obtained from these two techniques, indicating the accuracy of the automatic image processing algorithm. The image processing algorithm was then used to analyze pavement condition data from 450 surveyed miles of Salt Lake City street network. This API system resulted in significant savings in manpower needs while improving the consistency of the pavement condition data.