Research On Pavement Crack Detection Method Based On Machine Learning

With the rapid development of our society,the highway transportation industry also develops rapidly.As of the end of 2019,the total mileage of China’s highways has reached5.01.25 million kilometers.The year-on-year increase in highway mileage has also caused new problems-pavement maintenance.Traditional pavement condition survey and evaluation are based on manual work,which also has problems such as low efficiency,low reliability,and low safety.Considering these factors,an automatic road detection system is very necessary,and road cracks are one of the most common road diseases,so the research on road crack detection problems becomes the key.The field of image processing and machine learning has developed rapidly in recent years,which provides a theoretical basis for the research of this article.On this basis,this paper chooses the method of combining image processing and machine learning to realize the whole process of crack extraction,classification and road health assessment.Among them,a series of operations including median filtering,improved Canny edge detection,morphological processing,connected domain and skeleton extraction are carried out on the extraction of cracks,and the binary image of the cracks is successfully extracted.At the same time,for the extraction of pavement cracks under complex backgrounds such as water stains and rutting marks,a method based on probability statistics is proposed to complete the initial location of the crack area,and then image the crack area,which can also be completed for the road surface.Extraction of cracks.For the extraction of crack features,this paper selects 18 features based on geometric features,statistical features,and texture features,and uses principal component analysis to extract the first five principal components to realize the feature extraction of crack images.For the classification of cracks,the support vector machine method combined with the hierarchical method was selected to construct a complete binary tree,and the recognition rate of cracks reached more than 90%.Finally,combined with PCI pavement evaluation indicators,the evaluation of pavement health is completed.