The Study Of Pavement Cracks Automatic Identification Technology Based On 3D Laser Point Cloud

At present, with the rapid development of economic construction, the mileage of highway is increasing. When highway stimulates the development of various regions in life level of ascension, many problems will be found. Affecting by various natural and unnatural factors when they are used, roads can get a variety of diseases. Among these diseases, cracks are a common and harmful diseases. Crack is caused by the temperature, weather, road structure, the use of time and other various aspects factors, it will directly reduce the smoothness of roads and impact of vehicle driving. On the other hand, cracks will seriously affect the waterproof of roads, this will bring more serious effects for the road structure. Therefore, in order to improve the service life and security of roads, the damaged pavement maintenance has become a important topic of road maintenance management department.With the continuous development of science and technology, the research of automatic identification technology also constantly breakthroughs. Recently, the research of pavement crack automatic identification mainly revolves around gray images. The influence of real conditions of the pavement crack detection, such as illumination, oil and tire marks, hadn’t made the accuracy ideal. Therefore, this paper does the research of pavement crack automatic identification technology based on the data of high precision three-dimensional point cloud.Learn from the traditional image processing methods based on gray scale images, the number of different combinations of treatments has be applied to identify cracks in the pavement three-dimensional point cloud.This paper mainly aims at the three-dimensional point cloud of roads to complete some jobs of the following several aspects. Firstly, through a pavement three-dimensional point cloud acquisition devices which is combined with a deep camera and a laser transmitter to collect road point clouds. And a elimination method of zero value points for a number of zero value points in the cloud is used to eliminate these zero value points. Then the median filter and Gaussian filter respectively smooth the point cloud. And taking these two filtering methods which have their relevance into account, this paper also combined these two filtering methods to try to smooth the point cloud. Secondly, this paper analyzes the methods to describe the characters of crack points based on the height difference, normal vector, gradient and second derivative and studies feature extraction by using these methods of description. Finally, using genetic algorithm which uses the Otsu algorithm as the fitness function to calculate the optimal threshold to segment road background. And identify the target crack with the minimum spanning tree algorithm.The experimental results show that the automatic identification method of road cracks based on 3D laser point cloud in this paper is feasible. Cracks in the road point cloud can be identified by combining the methods of the preprocessing, feature extraction, point cloud segmentation and recognition.