Detection And Evaluation Of Pavement Distress Based On Quadtree Decomposition And Merge

Automatic pavement distress detection and assessment is a research hotspot of computer vision,because it can effectively improve the management and maintenance of pavements,resulting in economic benefits.But due to various pavement materials,colors, textures and light,the results of the assessment particularly affected.At present,road damage assessment methods must mostly people browsing through every piece of the road pictures,in the picture more or less marked the extent of the damage region and the extent of damage,and finally through a number of statistical methods to be the section of a comprehensive evaluation,with to determine the need for the refurbishment of the road. Although damage to the road edge detection method is light,texture and so is relatively sensitive,but more than image segmentation method is much less affected,if the original image to deal with a number of filter can effectively enhance the road damage to improve the accuracy of automatic evaluation.In this paper,the traditional path of damage based on the quadtree separation and merger methods and different types of damage to the type of experiments,experimental results show that the algorithm can be used not only cracks in the pavement damage category,the same can applies to crack or split type of road block damage and assess the accuracy of relatively high.From the original gray image and binary image,the result of quadtree splitting and merging,we get a 51-dimensional vector.Then using those vectors trains SVM and classify those images.This paper divides images into 3 main categories:no distress,linear distress and surface-like distress.Finally,different types of pavement distress are assessed statistics of length,width or area.The experimental results show that the detection method proposed in this paper achieves good results,I believe that through continuous efforts and improvements,the detection algorithm will have broad application prospects.