Automatic Identification Algorithm For Image Based On Asphalt Pavement Crack

With the rapid development of urban roads and highways,more and more people use asphalt concrete as pavement material.It is cause the road maintenance work become increasingly serious.Road maintenance work involves road detection,traditional methods can not meet the development of road maintenance.Automatic detection of new technologies and methods become the focus of research.So this paper research the automatic identification of asphalt pavemen image.Firstly in the introduction of background papers significance of the subject,current research,and then summarizes the structure of paper.On the basis of the history of introduction of asphalt pavement,pavement defects image characteristics,research methods asphalt pavement gray image processing,and using a weighted neighborhood average filtering method for image smoothing process.Then this paper,based on gray morphology filtering operator further enhance the image processing.This method can keep the target edge information while suppressing noise,and the effect is obvious.And then using edge detection operator to detect the image,according to the experimental test results,and ultimately determine the Sobel operator in the edge detection program.On the basis of mathematical morphology edge detection on the opening and closing operation method effectively empty and isolated point noise edges of the image have been processed.Finally,the object and the background image region segmentation process.Thesis focuses on the image after preprocessing feature selection and feature extraction: Disease extracted image feature selection method based on analysis of the disease and the characteristics of the image of the three statistical characteristics: The first is the image of disease in the horizontal and vertical pixel chart projection feature vectors extracted horizontal and vertical projection of feature vectors X_max,Y_max,And the number of connected components of the image feature vector N According to the basic theory of decision tree classifier,the feasibility study to analyze the experiment,Then classifier design and application identification sample classifier was trained.Finally,the experimental samples to verify the classification and identification of its final recognition rate reached 89.65%.The results show that the proposed method using the theory of pattern recognition methods to automatically identify disease feasible to achieve the purpose and significance of this paper.Pavement Distress image for automatic identification of research areas have some positive meaning.