Research On Automatic Pavement Crack Detection Based On Digital Image Processing

During the process of road maintenance, crack distress appear mostfrequently among all distress types. It has the greatest impact on the roadquality and safety. It is now the most importance indicator in the pavementevaluation process. At present, most of the domestic road maintenancedepartments mainly rely on manpower to handle pavement distress surveyneeds. Such way has several disadvantages such as large human powerconsumption, inaccurate results due to human interaction, traffic impactand safety problems. Thus based on digital image processing technology,pavement crack detection algorithm has made is possible for fast andaccurate automatic surveys.This thesis studies the key algorithms of the pavement crack detectionprocess and has covered the following contents:1) This thesis studies pavement distress image acquisitoin equipmentand points out that artificial lighting is the best way to have stable imagequality. And this thesis illustrates pavement crack charactistics.2) This thesis studies pavement image optimization technology,pointing out that pavement markings which is commonly used on currentpavements have great impact on further crack detection results. This thesispropose an algorithm based on histogram projection and lowess fittingalgorithms to eliminate such affection. This algorithm can not onlyeffectively remove the marking, and also preserve crack informations. Thisalgorithm also has small time consumption.3) This thesis studies crack extraction algorithms and comparesseveral commonly used algorithms. This thesis propose an algorihm basedon text recogonition algorithm and achieved good results. 4) This thesis studies crack measurement algorithms based on formerresults. This thesis uses graph morphology to smooth cracks and fill smallholes inside cracks. It distinguishs difference crack types based on holeareas. Then this thesis proposed a measurement algorithm based onimproved skeleton algorithms and graph theory. And based on the results,this algorithms measures crack width on original image.After a large number of image experiments, this thesis concludes thatthe crack detection process in this thesis can effectively identify andmeasure crack distress in pavement images.