Research On Pavement Crack Detection Based On Image Analysis

Nowdays, with the construction of highways, the research on road maintenance and pavement crack detection is being a focus. Because of the complex conditions of pavement, the existing detection techniques are generally artificial semi-automatic, which need numerous resources, are low timeliness and poor reliability. This makes the automatical pavement crack detection technology based on image analysis become a research hotspot.The problems are:numerous redundant datas, uneven illumination, serious noise and incomplete extraction. This research work is carried out according to these four aspects.(1) For the first problem, a pavement crack pre-detection method base on AMFT and projection feature is proposed. Experimental results show that this method can reduce a large amount of redundant datas processing.(2) For the problem of uneven illumination, the method of gray vertical normalization is used to balance the image illumination and the method of background fitting based on bilinear interpolation and multiplicative model is used to remove the image shadow, In addition, in order to avoid bluring the crack edges, the CESF model is used to enhance the gray value of edges of pavement carck before denoising process. The experimental results show that the methods can remove the impact of light while better retaining and even increasing the characteristic of pavement crack.(3) The research of image denoising algorithm is the focus of this thesis. In this thesis, the improved P-M diffusion equation based on8-neighborhood is proposed. Experimental results show that the improved method makes great improvement in image denoising. More important, the Shearlet Transformation is introduced in the denoising process of pavement image. The methods of corresponding adaptive threshold estimation and coefficient processing are proposed. The experimental results show that the methods can achieve good results in the aspects of noise smoothing and edges keeping.(4) For the last problem, a crack extension method based on local direction feature and gray value characteristic is proposed after the initial segmentation by FCM. The direction and the directional derivative factors are designed according to the local direction features. The local characteristic of crack is considered fully in the crack extension method. Experimental results show that this method can achieve better extraction of the pavement crack.