Research On The Algorithm Of Image Crack Extraction Based On Genetic Programming And Percolation Model

With the rapid increment of road traffic mileage in our country,the maintenance of highway has put forward a severe challenge.Crack is the main embodiment of highway disease.Automatic crack detection technology can overcome the shortcomings of traditional artificial crack detection,such as high cost,low efficiency and inconsistent detection standards,which is currently a research hotspot in crack detection technology.Therefore,it is great theoretical significance and wide application value to make use of crack detection technology to realize automatic testing of concrete pavement.The technology of the surface crack extraction of concrete pavement and its advantages and disadvantages at home and abroad were studied in this thesis.By analyzing the grayscale and shape features of concrete road surface cracks,further researches on the difficulties of extracting cracks by existing algorithms are made.The main research work as follows:1.Summarize the related literature of crack detection technology at home and abroad.Combining the gray features of the crack image on the surface of the concrete pavement,the light normalization treatment of the concrete pavement crack image was realized.The problem of slow convergence of crack detection algorithms based on genetic programming was studied.By improving the genetic selection operator,the training time of the algorithm was accelerated while the accuracy of the algorithm was guaranteed.2.On the basis of the detection results of genetic programming algorithm,the improved Zhang parallel thinning algorithm was used to thin the image.Then a burr-free crack skeleton with a single pixel width is obtained.According to the characteristics of the skeleton endpoints,the endpoints of the fractures were calculated.Research and improve the percolation algorithm,and directly use the end point of the fracture skeleton as the anchor point to conduct the seepage detection.3.In-depth study of the interference of cracks extraction in the background area of concrete highways such as surface pits,seepage water,etc.According to the characteristics of the surface image of the concrete highway,this thesis studies the noise removal algorithm based on the characteristic of the connected region.The improved connection algorithm was studied to locally refine the fracture cell intersections and remove the erroneous forced connections,thereby improving the accuracy of crack detection.4.The extraction algorithm of concrete road surface cracks based on genetic programming and seepage model was studied.And the proposed algorithm compared with the current mainstream algorithm performance and detection accuracy.To verify the effectiveness of the proposed algorithm.Based on the research of image crack extraction algorithm based on genetic programming and percolation model,a complete,efficient and high-precision concrete surface crack extraction system was developed.The system can realize the automatic detection of cracks which meet the actual needs of the project.Finally,this research work had been summarized and the future prospect had been pointed out.