The Research And Implementation Of A Pavement Crack Automatic Detection Algorithm Based On Multi-scale Characteristic Analysis

Crack disease is not only the main form of road diseases,but also an important part of road detection.Due to the complexity of the pavement environment,the existing crack detection algorithms are usually inefficient or not universal,and it is often difficult to achieve the expected result.Therefore,it is urgent to improve and optimize the existing detection algorithms.In view of the above situation,based on the analysis of the existing literature,this paper makes an in-depth study and software realization of the initial location of pavement crack area and the high precision detection algorithm of pavement crack.The main contents are as follows:(1)An initial location algorithm for pavement cracks based on deep learning is designed.Firstly,based on the brightness elevation model,an improved dodging algorithm is developed to eliminate the uneven distribution of pavement image brightness.Then,the pavement images are diveded into sub-blocks,and the improved Lenet-5 convolution neural network model is trained by marked non-crack sub-blocks and crack sub-blocks,so as to realize the initial positioning of the crack sub-blocks in the pavement image.(2)A linear crack detection algorithm based on multi-scale ridge edge is designed.Through the analysis on the ridge edge characteristics of pavement cracks,the principle of ridge edge detection based on the first derivative and second derivative of each point is obtained.Then,based on the multi-directional and multi-scale characteristics of the crack,a bidirectional and multi-scale fusion detection algorithm is designed,and the fake cracks are removed by mathematical morphology and connected domain algorithm.(3)A meshy crack detection algorithm based on Hessian matrix is designed.Based on the analysis of the Hessian matrix eigenvalue,the multi-scale linear filter is constructed according to the different characteristics that the eigenvalue shows on different twodimensional geometric structure.The linear response intensity of each point in the pavement image is obtained by filtering,and then the image is binarized according to the response intensity and the fake cracks are removed.(4)A pavement distress evaluation algorithm is developed.Firstly,a crack connection algorithm based on the minimum spanning tree is proposed to enhance the continuity of crack segmentation results.Then a set of crack geometric parameter detection algorithm is developed to detect the binary image.Finally,the pavement distress is evaluated by the detected parameters.This paper designs a series of important algorithms for pavement crack detection from image preprocessing to damage assessment.The algorithms make full use of the multi-scale and linear feature of cracks,and have strong adaptability to uneven brightness and noise.Through a large number of experiments,the proposed algorithms are superior to the classical detection algorithms for comparison of detection accuracy,noise immunity and versatility,and has certain engineering and promoting value.