Research Of Pavement Crack Detection Algorithm Based On Image Processing

Since the 21st Century,China's economy has made rapid development.If a country wants to be prosperous,its transport should be well-developed.Therefore there has been a large-scale expansion of highway construction.At the same time,the growth of area has also brought the problem of pavement maintenance work.Most of the pavement damage form is pavement cracks.If the road cracks can be repaired timely,it will save a lot of manpower and financial resources.In recent years,intelligent detection technology of pavement cracks has been widely used.But because intelligent detection technology in China has been late development and it is still not perfect.So its precision is not high enough.The main research object of this paper is the intelligent identification technology of pavement cracks.Because pavement crack information and pavement background contrast in the original pavement crack image collected is relatively low.The paper will pre-process the image first,mainly including the pavement crack image intensity,enhancement and noise-removal handling.This paper proposes an improved Retinex algorithm to enhance the crack image,which can provide enhanced pavement cracks and the background contrast more intuitively.In the pavement crack image segmentation,this paper selects two kinds of algorithms,which is 8 direction based on Sobel algorithm and segmentation algorithm based on nearest neighbor mean shift image segmentation algorithm.Due to the irregular pavement cracks and impurities,the first algorithm effect is not ideal.Many isolated noise points are not removed,and the second algorithm will also remove the majority of the interference noise and improve segmentation accuracy in segmentation of the crack.Although the segmentation result of the first algorithm is more complete and more favorable in the extraction of fracture characteristics of pavement cracks,there are still some isolated noise points.The feature extraction of this paper adopts morphological dilation erosion operation.Not only does it extract the crack shape,also eliminates the isolated noise points,which lay a solid foundation for later identification.In recognition of pavement cracks,whether it is linear cracks or net cracks in the background image based on the number of connected domain.Whether it is the transverse cracks or longitudinal cracks depends on the linear projection of the X and Y axis.All the algorithms in this paper are verified on the MATLAB platform,and most of them are proved to be satisfactory.It can judge pavement program information according to the pavement damage degree of PCI while in the parameters of the identified crack(crack lengths,crack width,crack area extraction,etc.)To a certain extent,it improves the efficiency of road detection technology and provides road maintenance department with more direct information.