Feature Extraction Method Of Highway Surface Crack Image

Highway pavement crack detection system is in a moving vehicle, the real-timecollection for the crack of asphalt pavement, and the acquisition of informationprocessing, analysis and preservation, an information detection system provides trafficinformation for road management department. Management personnel according tothe obtained information to develop the highway maintenance planning, is one of theimportant part of highway management system.Extraction of pavement crack is the main task of automatic crack detection system.Among them, pavement crack extraction technology determines the level of accuracyof judging the results make the highway pavement condition. Extraction methods ofpavement cracks of the digital image processing technology, is widely used inautomatic crack detection system. However, due to the crack shape with a high degreeof uncertainty, quantitative criteria for crack image extraction method, the existingdoes not have a unified, easy to cause differences in crack space structure andevaluation details. In this paper, in view of the above questions, research from t heaspects of image enhancement, image segmentation and crack feature extractionmethods, the main research work includes:(1) the image preprocessing. Introduces the related knowledge of image denoising,image enhancement, image denoising introduces Gauss noise, wavelet denoising;image enhancement introduces linear point operations, non-linear point operation,puts forward an improved wavelet multi-scale image preprocessing algorithm.(2) image enhancement algorithm. Improvement of the two algorithms, abackground gray processing based on statistics, another is the background estimation,the two improved methods are simulated by Matlab, the calculation results show thatthe two algorithms can effectively separate the crack and background. (3) the crack image processing algorithm. In the crack image processing algorithms,proposes a maximum posterior probability measure based feature extraction algorithmcombined with multiple cracks, the algorithm uses the gray information of feature andedge histogram characteristic crack image as the characteristic crack image model,using the maximum a posteriori probability index as the crack similar centroiditeration method of measurement for to extract the crack characteristics. Through theanalysis and simplified on gray characteristic crack image, makes it possible tocombine feature has better resolution performance and lower complexity, at the sametime, in order to solve the problem of crack model drift model from the update brings,proposed a model update strategy.(4) the simulation experiment. On this crack image preprocessing algorithm and thecrack image feature extraction algorithm is validated by simulation. The simulationresults are presented, and the measurement index of crack characteristics is given.