Asphalt Pavement Damage Condition Evaluation Applicable To Multiple Types Of Interference

In recent years,with the rapid development of China’s economy and society,the travel demand of highways has grown rapidly and the scale of road maintenance and repair has expanded.But the pavement damage detection exposed to the natural environment faces interference from various factors.Existing detection algorithms are sensitive to the interference conditions,which makes it difficult to carry out extensive applications in actual detection.In addition,early repaired damage areas are starting to show secondary damage due to corrosion and wear,but there is a lack of detailed studies on secondary damage detection in this field.This thesis proposes a pavement damage detection and evaluation algorithm that is more suitable for practical applications.It complements the research on secondary damage,and designs countermeasures for each type of interference encountered in practical inspection.The purpose of this paper is to ensure the detection accuracy while suppressing the acquisition equipment interference,road structure interference and natural environment interference in the actual engineering detection as effectively as possible.The main research contents of the thesis are as follows:1)Raster correction algorithm based on longitudinal grayscale template learningIn the actual road inspection project,because the acquisition equipment is affected by electrical or electromechanical factors,the images on the output end of acquisition will have a longitudinal distribution of gray scale uneven raster stripes.These raster stripes throughout the area to be detected,not only will affect the accuracy of the segmentation of the damage,and even confuse with the real repair of the damaged area.Therefore,this thesis analyzes and summarizes the distribution pattern of raster stripes on the road data set,then derives and calculates the raster correction vector and raster correction matrix.We use a raster correction algorithm based on longitudinal gray-scale template learning to suppress or even eliminate the raster effects caused by the acquisition equipment,so as to avoid interference with the subsequent detection.2)Automatic crack detection based on cooperative filter groupSince the detection of fine and shallow cracks under low contrast conditions is extremely difficult due to the varying structural scales,this thesis incorporates multi-scale features on the basis of traditional filters to enhance the response of fine and shallow cracks.In order to solve the problem that the branching and bifurcation points are easily broken in the crack extraction process,two sub-filters are designed and added on the basis of the main filter for co-detection.The multi-scale offset features are incorporated between the co-filters to solve the problem that the branching and bifurcation points are easily missing in the crack segmentation.For the phenomenon of obvious inversion of the gray value features of crack-type breakage under different wear and corrosion conditions,two corresponding sets of co-filters are designed in this thesis for efficient and accurate detection while effectively suppressing road texture noise.3)Adaptive repair segmentation algorithm based on contrast classificationA clustering algorithm is used to analyze the distribution of gray values in the input road image to determine the type and contrast characteristics of the pavement repair.Using the segmentation algorithm based on the maximum extremal stable region to extract the repair-type damage under high contrast conditions,and using the multi-node growing algorithm based on super pixel segmentation to segment the repair-type damage under low contrast conditions.The initial selection of the candidate region set is based on the area characteristics and relative gray scale characteristics of the regions.4)Interference discrimination algorithm based on feature fusion selectionThe actual road inspection project faces a wide range of interference situations,and the interference area is often similar to the real damage form.But there is always a certain degree of difference.In this thesis,we classify and summarize the common types of interference in road inspection engineering,and use a feature fusion selection based approach to select candidate damage regions.A feature fusion-based selection method based on the individual characteristics of each interference is proposed to select candidate damage areas.Then we design discriminative strategies and selection schemes corresponding to real damage and interference areas.5)Automatic pavement condition evaluation algorithm based on multi-feature fusionIn order to evaluate the damage condition of the road in a timely manner,this thesis designs an automatic pavement condition evaluation algorithm based on multi-feature fusion to evaluate the type of road damage and the level of damage.Firstly,the damage area is coarsely classified into linear and reticular types based on the projection features and basic geometric features of the crack area.Secondly,the damage distribution density of the crack area and the number of major crack blocks are calculated to sub-classify the crack types.Finally,the crack skeleton features are extracted and the width of the cracks are calculated.The type and damage level of the road damage are judged comprehensively by combining the area of the major crack blocks.The algorithm in this thesis is tested on the DMU-Damage dataset.The DMU-Damage dataset is collected in actual road inspection project which contains all the common types of interference and damage.The experiments show that the proposed method in this thesis can effectively suppress the influence of interference while accurately extracting the real damage of the road.So it can be put into the actual road inspection project in use.