Research On Pavement Crack Detection Algorithm Based On Digital Image Processing

With the rapid development of the transport industry,China’s traffic flow continues to increase,the use of the road surface is also on the rise,due to the temperature of the environment,humidity and applied loads and other factors,road pavements are subject to damage such as cracks,rutting and subsidence and other damage phenomena,of which cracks is one of the initial manifestations of pavement damage,and an important indicator of the pavement,when the cracks are damaged to a certain extent,it will bring traffic hazards.At present,the common means of pavement inspection is manual inspection,which is inefficient,inaccurate,time-consuming,and subjective,and affects traffic.With the rapid development of image acquisition technology and computer technology,digital image processing technology has been widely used in the field of pavement crack detection.The steps of pavement crack detection based on digital image processing are: firstly,a vehicle-mounted camera is used to capture pavement cracks and then algorithms are used to segment and classify the cracks.However,the existing methods for pavement crack detection usually focus on only one task in crack segmentation or classification,lacking a comprehensive research to achieve segmentation and classification of pavement crack images at the same time.The current detection algorithms for pavement crack images can be divided into traditional algorithms and machine learning algorithms,and with the rapid development of artificial intelligence technology,the accuracy of deep learning algorithms in machine learning algorithms for crack detection is gradually increasing,based on this,this thesis uses digital image processing technology and deep learning algorithms to realize the segmentation and classification of pavement crack images,and automate the calculation of pavement deterioration rate and pavement condition index,and the specific tasks are as follows:（1）To address the problem that cracks occupy a relatively small portion of crack images and are difficult to segment,this thesis proposes a pavement crack image segmentation method based on the combination of U-Net and residual network（Res Net）,i.e.,the organic combination of U-Net and Res Net to construct a U-Res Net network,which can achieve the effective fusion of high-level semantic features and low-level semantic features,so that it has strong representation learning ability on cracked pavement image data;by adding the residual network,the network also has good robustness in segmenting cracked images with small data sets.The experimental results show that the proposed method is robust and has good generalization capability.（2）To address the problem of calculating different deterioration rate for different pavement crack types,in this thesis,the segmented binary crack images are further classified to enable automatic calculation of parameters related to pavement condition,and proposes a two-branch collaborative constraint network-based pavement crack image classification method,i.e.constructing a collaborative constraint network with Res Ne Xt as the backbone network,and in order to fully express the crack features,the segmented binary image and the original color crack image are used as the dual-branch input in this thesis.The experimental results show that the proposed method can effectively identify the crack categories of the segmented images and automatically calculate the pavement deterioration rate and pavement condition index according to its classification effect.