Research On SVM—Adaboost Crack Classification Based On Morphology

In recent years,many researchers have proposed various research-programs and tried to nondestructively extract as much as possible the cracks of the road pavement in the process of road surface crack image detection,especially the cracks around the trunk cracks.However,Pavement cracks themselves contain irregular textures,coupled with the interference of complex information such as grunge,shadows and noise,so nondestructively extracting the cracks is still a significant and challenging subject.In the real life,road image capture is often affected by the external environment,and there exist different degrees of noise,which caused interference in the image processing process.Based on these interference,the algorithm of crack image processing is given.In this paper,through researching the various steps of traditional image processing and comparing the experimental results of different algorithms,and point at the shortcoming in the common algorithms,the algorithm was improved.In the image segmentation stage,an improved DBC segmentation method was proposed.Through the experiment test and effect comparison of the DBC method,it was found that the improved algorithm can obtain the ideal image segmentation effect in the experiment.At the same time,for the extraction of images with complex fractures,this paper proposed a ridge-threshold crack detection algorithm based on fractional integral.Through the comparison and analysis of experimental results,the generalization and rationality of the algorithm was proved.In the classification stage of cracks,although the recognition rate of SVM classifier is higher,the classification effect of longitudinal crack and block crack is not ideal,and the applicability to the real road is poor.While the algorithm Adaboost can adaptively adjust the distribution of training samples,thereby improving the classification accuracy.Based on the above advantage and disadvantage and SVM—Adaboost image classification algorithm,this paper achieved the correct classification of the image by using the machine learning.