Research On The Detection Of Pavement Crack Based On Image Processing

Road traffic plays an important role in the field of transportation in China.With the continuous development of the economy,the traffic volume of the road is rapidly increasing for more and more vehicles.The pavement will be inevitably damaged to a certain extent,in which cracks are a common hidden security danger.At present,a series of algorithms are proposed to improve the efficiency of pavement crack detection for highway pavement diseases at home and abroad.Compared with the highway surface,the cracks in the airport pavement are smaller and the detection requirements are more stringent,so the requirements of airport crack detection cannot be met by the existing algorithms.Currently,most airports still use manual observation methods.Based on the detection of road surface cracks of artificial observation,there are several problems: 1)Human observation is a labor-intensive work,which is more time-consuming;2)the test results are difficult to repeat,lacking of uniform testing criteria;3)it is difficult to ensure that the detector can cover the entire airport runway.Therefore,there is an urgent need for an automated algorithm for airport crack detection.In this paper,based on image processing,an algorithm is proposed to solve the problem of airport crack detection,which consists of three aspects:(1)In order to improve the accuracy of surface crack detection and overcome the interference of noise of airport channel background,a crack localization algorithm based on deep learning is designed.The algorithm includes three parts: the selection of deep learning framework,the generation of datasets and the training of models.Firstly,the image can be divided into two categories by this algorithm: crack image and non-crack image,and the crack area in the crack image is labeled as the input of the next crack pixel detection,which narrows the detection range.(2)The crack is found in the rectangular area through the deep learning of crack localization algorithm as the first step,the next step needs to find out the localization of the crack pixel,this paper presents a road surface crack detection algorithm based on the shortest path spanning.The algorithm includes three parts: seed point extraction,shortest path spanning and growth algorithm.It compares with the existing crack algorithm by testing in the exposed dataset AigleRN and CFD,and the results show that the detection result of this algorithm is better than the existing algorithm,the accuracy is higher and the calculation is simple.By detecting the results of airport pavement data,it is faster than detecting the whole image directly,and the crack position detected is more accurate.(3)In order to have a comprehensive understanding of the surface disease,this paper presents an image mosaic algorithm for the visualization of road surface cracks.This algorithm includes: Image illumination compensation algorithm,image initial mosaic algorithm based on location information,stitching optimization based on feature point matching.The algorithm finally realizes the mosaic of nearly 6,000 airport pavement images of an airport in South China and an airport in Southwest China.Finally,the algorithm is tested on several airport and open highway datasets,and get better results.Therefore,the algorithm in this paper is also applicable to crack detection of highway pavement.