Research On Asphalt Pavement Pothole Detection Technique Based On 3D Line Laser Technology

As the main damage type of asphalt pavement,potholes not only affect the road structure,but also harm for the traffic safety.With the continuous growth of highway mileage and the rapid increase of traffic volume,the task of highway rehabilitation and management is heavy and urgent.A comprehensive and reliable method of acquiring pothole properties can not only prevent traffic accidents effectively,but also provide a foundation for rehabilitation decision-making.It is important to ensure road capacity and service level,improve the scientific level of road maintenance decision-making and the efficiency of rehabilitation funds.In China,the severity of potholes is classified by its area properties.However,the depth of pothole varies due to the complexity of pothole which means that different damage degrees of pavement structural layers maybe have the same area of pothole.Obviously,it seriously affects the identification of pavement structure damage layers.Meanwhile,it fails to reflect the real interaction between the pavement and vehicles.Therefore,the analysis of existing potholes area data unable to provide comprehensive and accurate potholes information for managers,which restricts the scientific and refined development of highway maintenance.This paper firstly explains the composition and principle of the three-dimensional laser detection system in details.And a 3D line laser detection system and a detection vehicle are constructed.During the detection process,we find out some missing points in the original data.Therefore we develop a eight-neighborhood filling method and threshold elimination method to process the data.Secondly,this paper proposes an extraction method of pothole properties which has been validated in laboratory test.Besides,the influence of adjacent spacing between two laser profiles on the error of pothole properties is analyzed.And regression analysis has been used to predict the relative errors based on different spacing.In the practical environment,convolution neural network algorithm has been used to identify potholes.In order to improve the efficiency of pothole properties measurement,image binarization and oriented bounding box have been used,so that extraction method of pothole properties can concentrate on pothole feature region.Finally,this method also be verified in the field test.Based on MATLAB software,a programming for extracting pothole properties is developed.Based on 3D line laser technology,the method of pothole identification and pothole properties extraction is conducive to the existing potholes detection method in China.This research is contribute to improving the detection method of of asphalt pavement distress.And it is essential to the scientific and refined development of highway maintenance.