Research Of Pavement Crack Image Recognition Algorithm

The development of the pavement in our country is rapid.Seasonable pavement quality testing can not only prolong the road service life,but also avoid the hidden trouble of vehicle driving which comes from the pavement disease.Thinking about that traditional detection method based on artificial test exists a series of defects such as poor efficiency,low accuracy,high risk about safety and so on,then the pavement crack automatic detection system has become the front-burner research issue.As well as,the core content of the automatic detection system is crack automatic detection algorithm.According to the existing domestic and foreign algorithm of pavement crack detection,this thesis dose the research and design about the algorithm of pavement crack detection.Firstly,traditional pulse coupled neural network(PCNN)model was simplified if just form the perspective of detecting cracks simply in the crack images.Simplified PCNN model not only can reduce the computation complexity of traditional PCNN model's simulation,but also retain the operational characters of the original neural.For PCNN model is not sure of the optimal detection of the crack images and pulse threshold with nonlinear factor,a method of crack detection of crack images is proposed based on genetic algorithm and optimized PCNN model—GA-PCNN.The method is on the modified minimum error principle as the fitness function of genetic algorithm,and according to the characteristics of the genetic algorithm has the global optimal solution to determine value of each factor in simplified PCNN model to realize automatic segmentation of simplified PCNN crack images.A kind of multi-structural elements anti-noise edge detection operator was put forward which was based on mathematical morphology arithmetic to extract the target crack after using the algorithm of GA-PCNN to segment the crack images.Then uses based on the growth of the active method of connecting the fracture cracks piece to deal with the wider fracture cracks.The proposed algorithm in this thesis is simulated under MATLAB R2009 a platform,meanwhile,comparing the processedresults with other methods of segmentation come from the road crack images,quantitative analysis is conducted for the image after segmentation using region contrast,ROC curve.Experiment results show that the proposed crack detection method is effective and universal.Finally,extracting,classifying and calculating the feature information form the result image which has been detected with the above method.By supposing a set of conditions,the connected domains information of cracks can be extracted,as well as the presented characteristics of the crack projection can be observed,and then judgment the type of cracks.In the end,calculate the target crack features' values such as area,length and width by getting cracks skeleton with the method of refining.