Development of Digital Image-Processing Algorithm to Compute Unified Crack Index for Salt Lake City

Abstract

This paper presents the recent efforts in developing an image processing algorithm for computing a unified pavement crack index for Salt Lake City. The pavement surface images were collected using a digital camera mounted on a van. Each image covers a pavement area of 2.13 m (7 ft) × 1.52 m (5 ft), taken at every 30-m (100-ft) station. The digital images were then transferred onto a 1-gigabyte hard disk from a set of memory cards each of which can store 21 digital images. Approximately 1,500 images are then transferred from the hard disk to a compact disc. The image-processing algorithm, based on a variable thresholding technique, was developed on a personal computer to automatically process pavement images. The image is divided into 140 smaller tiles, each tile consisting of 40 × 40 pixels. To measure the amount of cracking, a variable threshold value is computed based on the average gray value of each tile. The program then automatically counts the number of cracked tiles and computes a unified crack index for each pavement image. The crack indexes computed from the image-processing algorithms are compared against the manual rating procedure in this paper. The image-processing algorithms were applied to process more than 450 surveyed miles of Salt Lake City street network.