Research On Image Crack Edge Detection Method Based On Percolation Model

With the increasingly perfection of traffic road construction in China,the traffic system is more and more concerned about the detection and maintenance of highway infrastructure.The degree of road surface damage has seriously affected the operation of roads.The crack is a common form of damage.If the crack can be detected and repaired in time,the safety of traffic could be guaranteed and the service life of traffic could be prolonged.Therefore,the research on intelligent detection methods for pavement crack based on digital image processing has certain practical significance.The existing crack detection algorithm is mainly for the complete extraction of crack targets.And the actual application requirements of engineering often do not need high quality and high resolution crack images.So the edge detection method of road pavement crack was researched in this thesis.The characteristics of road pavement crack were analyzed from the global and local characteristics of crack images.And the key steps in the process of crack edge detection,such as edge detection,and crack target extraction were studied in the thesis.The edge detection method for road pavement crack based on percolation model was proposed.In the method,an adaptive threshold method was designed to extract the candidate points of the crack.And based on percolation model and the gradient information,the crack edge detection method was proposed.The mainly research work of the thesis was carried out as following:1.The traditional crack detection and edge detection methods were studied in depth.Based on the preprocessing of dominant features enhancement of road pavement images and the global and local features of road pavement crack images,an adaptive algorithm for extracting candidate points of crack edges was proposed.According to edge point gradient information,road pavement crack images were segmented by adaptive thresholding.Then,the candidate points of the crack edge could be extracted and the main body in the crack edge area was rapidly located.2.Based on the thought of percolation model,the similar adjacent intensities of crack edge and grayscale difference between crack edge and background were considered comprehensively to improve the clustering feature conditions.Then the crack edge was detected in the thesis.In order to solve the problem of low detection efficiency for percolation model,a crack edge detection method based on crack edge candidate points extraction was proposed.The efficiency and accuracy of detection were raised.Then based on the extraction of crack edges,the complete extraction of crack targets could be further achieved through morphological operations.3.The edge detection method of road pavement crack based on percolation model was studied.A system of complete road pavement crack edge detection was designed in this thesis.The system mainly included image loading,crack image preprocessing,the extraction of crack edge candidate points,crack edge detection and the result output and storage.Experiments showed that the system could not only process in real time and meet the basic needs of engineering practice,but also achieve rapid and accurate location of crack targets.