Detection Of Highway Shallow Defect Based On Support Vector Machine (SVM) And Ground Penetrating Radar (GPR)

Defect existed in shallow roads not only affect the transport performance,but also reduce the life of the road.The shallow defects have certain concealment in the highway structural layer.But now,the conventional detection methods in China have many problems such as low detection efficiency,incomplete evaluation results,and heavy reliance on manual inspection.Therefore,an efficient and accurate automatic recognition system is in need of a development.This paper proposes a method for identifying shallow defects in highways based on the pattern recognition and the ground penetrating radar.The main research work of this paper is as follows:(1)By comparing the merits and flaws of core sampling,ultrasonic testing and ground penetrating radar,it is found that the efficient,accurate and non-invasive ground penetrating radar technology is the best research tool in this paper.(2)This paper introduces the basic principle of ground penetrating radar,the relevant equipment and software,the propagation principle of electromagnetic wave,the imaging technology of ground penetrating radar and the causes of common defects and discusses and analyses the imaging characteristics of cracks,voids and subsidence in ground penetrating radar.(3)By analyzing the merits and flaws of median filter,mean filter and gaussian filter during the research of image preprocessing technology,the effective gaussian filter is selected as the denoising algorithm,the Canny operator is used as the edge detection algorithm and the image morphological processing algorithm is used to remove information outside the target,and finally the target image is segmented by projection segmentation algorithm.(4)The original features such as area,complexity,texture features and seven invariant moments are extracted from the defects image,and the original features are compressed and dimensioned by K-L algorithm in order to reduce the computational complexity,reduce data overlap and improve data validity.(5)In this paper,support vector machine(SVM)classification algorithm is used to classify and recognize different defects.In order to solve the problem of parameter optimization of SVM algorithm,the grid search algorithm and particle swarm optimization algorithm are used to optimize the parameters of SVM.The classification and prediction results of SVM model obtained by these two optimization algorithms are 88.333% and 86.667%,respectively.By comparison test,it is found that the grid search algorithm has high accuracy,but it runs slowly and the speed of particle swarm optimization has been improved,but it is easy to fall into the local maximum.Aiming at the premature problem of PSO algorithm,the author proposes an adaptive mutation PSO algorithm by referring to the mutation idea of genetic algorithm,which makes the particle swarm can jump out of the local maximum,and then get the global maximum.The classification prediction result of the improved SVM model is 91.667%,which improves the recognition rate of the drawbacks.