Trr No. 1311 Pavement Design, Management And Performance, pp. 120-130
Digital image processing and pattern recognition from a video log of pavement surface distress is accomplished by initializing the image prcessing board within an IBM-PC (or compatible). Software is used to set all image processing registers to zero or a default value, which digitizes values for each color (0 to 255), and sets an image processing board to continuosly display signal readings from video log tape. The video signal is received by the image processing board and display on an image monitor. A pixel can represent 256 possible shades of gray, with 0 representing black, 256 representing white. The object of image processing is to enhance pavement surface distress (changing it to white) while diminishing the nondistress area (changing it to black). Pavement surface distress is distinguished by using one of several types of distress (crack) enhancement filters. For crack recognition, multiplication filters are useful, but multiplication filters cannot recognize crack pattern or classify type of cracking. Directional filters are used to classify cracking by splitting the image into horizontal and vertical components. The extent of cracking or crack width can also be determined by directional filters. Once the right filter value is selected, crack size and length are validated. The ARIA TM image processing technology is accurate, consistent, and reliable, and may be used on any type of concrete or asphalt pavement.