Identification Method Of Cement Pavement Clearance Based On Deep Learning And Ground Penetrating Radar

Cement pavement is affected by vehicle dynamic load,temperature gradient and rain penetration for a long time,which is easy to form the hollow disease between the cement concrete plate and the base.Under the action of vehicle dynamic load,it is easy to appear plate fracture,crushing and other more serious damage,affecting driving safety.If it can be found and solved in the early stage of the road without damage,the service life of the road can be extended,so it is urgent to establish the detection method of the road.Aiming at penetrating disease difficult to detect,Ground Penetrating Radar(GPR)is used to nondestructive test cement pavement penetrating disease.In order to obtain the characteristics of the unloading area,the spectral characteristics of the unloading distress of cement pavement were obtained by gpr Max forward simulation method,and the effectiveness of the spectral characteristics was verified by comparing with the results of indoor simulation and field experiment.The pretreatment method of GPR signal of cement pavement is studied,and the post-treatment process of GPR is obtained,which lays the foundation for obtaining clear GPR map.Aiming at the difficulty of GPR data interpretation,an improved Mask-RCNN recognition model based on feature fusion and attention mechanism is proposed.The GPR data set of cement pavement was constructed,and the training set and test set were divided according to the ratio of 4:1.The experimental results show that,compared with the original Mask-RCNN and father-RCNN,the m AP of the improved Mask-RCNN model reaches 90.9%(Io U=0.5),indicating that the proposed improved Mask-RCNN model has the highest identification accuracy and is more suitable for automatic identification of the cement pavement hollow-out disease.Aiming at the problem that the current GPR data preprocessing software is in the commercial black box state,a MATLAB based GPR data post-processing and void disease automatic recognition software is designed.Through MATLAB,the improved Mask-RCNN model is integrated into the software to realize the integration process of data processing and the identification of void disease,and the function experiment of the software is carried out.The experimental results show that the designed software can realize the pre-processing of GPR data and the identification of the unloading disease after processing,and realize the automatic identification of the unloading disease of cement pavement.The GPR signal processing software and the de-cavitation disease identification model studied can provide scientific decision-making basis for accurate pre-maintenance and safe operation of cement pavement.