Research On Highway Pavement Crack Recognition Technology Based On Image Processing

Pavement cracks are one of the manifestations of the initial stage of road surface damage.Accurately detecting and identifying pavement cracks,formulating conservation methods scientifically and reasonably,and preventing deeper damage to the road surface are important for reducing the national economic losses and ensuring people's travel safety.The traditional way of manual detection of pavement cracks can no longer meet the development and requirements of modern road maintenance.The research on the automatic identification technology of pavement cracks has been inescapable.In this paper,some problems in pavement crack identification technology are studied by using the theory of fractional differential,fuzzy logic,edge detection,threshold segmentation,mathematical morphology and machine learning.First of all,for the characteristics of road surface cracks such as uneven illumination and poor contrast,the traditional image enhancement algorithm is easy to ignore the details of crack texture,and it is easy to produce noise.This paper focuses on research and proposes an image enhancement based on fractional differential and fuzzy set theory.Using the functional relationship between image gradient and fractional differential,an adaptive fractional differential algorithm was proposed to preserve the texture details while highlighting the edges of pavement cracks.Combined with the human visual characteristics,the improved fuzzy set theory was applied to the pavement background effectively.Denoising,the organic combination of the two has a good effect on the pretreatment of the road surface image.Secondly,on the basis of image preprocessing,this paper carries out several experiments of edge detection and threshold segmentation on the road surface images respectively.The maximum class-to-variance method is used to compare the road surface image segmentation.Mathematical morphology is used.Using four basic operations in combination,the isolated noises after segmentation are removed,and the cracks are extracted and refined.Finally,the projection pixels of pavement cracks in the horizontal and vertical directions are extracted,and the total number of crack pixels is taken as the feature value of the pavement crack image.Using the “one pair of rest” classification method,the SVM classifier is constructed using the radial basis as the kernel function.The classification and recognition of pavement crack images are carried out.Experiments show that the pavement cracks are characterized by feature projections and the radial basis kernel function is used to classify the pavement cracks.