Research On Pavement Distress Detection And Classification Algorithm Based On Semi-supervised Learning Algorithm

Traditional pavement distress detection and classification method based onhuman’s visual sense has been difficult to adapt to the fast development of modernroad network scale, because it’s time-consuming, higher labor-intensive, lowercost-effective, dangerous and it may impact traffic. In order to overcome thelimitations of traditional pavement distress detection and classification method, andto meet the requirement of the modern pavement maintenance and management,many attempts recently have been made to achieve intelligent pavement distressdetection and classification system, so pavement images preprocessing, distressdetection and classification have become the research hotspots in the field ofhighway traffic technology and pattern recognition.With the image processing and pattern recognition technology, this papercompleted pavement images preprocessing, distress detection and distressclassification algorithms. The concrete research contents are as follows:(1) Explore a suitable pavement images preprocessing algorithm. As thepavement images are more complex than other images, the conventional imagepreprocessing algorithms are difficult to highlight pavement distress information, soit is necessary to explore an effective pavement images preprocessing algorithm.(2) Extract pavement images’ features and explore an effective pavementdistress detection algorithm. On the basis of images preprocessing, extractpavements’ features which can represent pavements’ characters. And discuss thedistress detection algorithm based on Radon transform.(3) Research the intelligent pavement distress classification algorithm based onsemi-supervised learning algorithm, and compares the performance withclassification algorithm based on Support Vector Machine (SVM). As the quantity ofsome distress samples is very little, and the semi-supervised learning method is goodat achieving classification with a few of marked samples. So it is necessary to try toexplore the intelligent pavement distress classification algorithm based on semi-supervised learning algorithm.The pavement images data in this paper come from Jiangxi TianChi HighwayScience and Technology Development Company. The pavement image preprocessingalgorithm used in this paper can effectively highlight the pavement distressinformation. The detection algorithm based on Radon transform is reliable, itsaccuracy can reach97%. The performance of classification based onSemi-supervised Learning and Support Vector Machine (SVM) were compared, andthe result shows the performance of classification based on Semi-supervisedLearning is better.