Study On The Identification Method And Maintenance Decision Of Rural Road Asphalt Pavement In Henan Province

With the rapid development of urban and rural economic integration and rural urbanization process,the rural road network system has been gradually improved.How to detect quickly and accurately the pavement diseases and formulate scientific and effective maintenance decision methods has become the key to the current rural road maintenance work.However,there are still few studies on rural road surface disease identification methods and maintenance decision.The previous road surface disease identification methods mainly rely on manual inspection and measurement,mainly rely on experience for maintenance,which consumes a lot of manpower,material and financial resources and has problems of low efficiency and low accuracy.Based on this,this paper takes the rural road asphalt pavement,studies the pavement disease identification method,and makes a scientific and effective maintenance decision.The contents of the study mainly include:(1)According to the classification and technical standards of rural roads and the actual investigation of rural roads in Henan Province,we summarize four typical types of road diseases,including cracks,loose,deformation and others,analyze the causes of different diseases;loss function,single-stage target detection algorithm and two-stage target detection algorithm,which lays a foundation for the following research.(2)Establish a high-definition image data set of a rural road in Henan,Create the experimental environment and the evaluation indicators,Using SSD,Faster RCNN,YOLOv3,YOLOv4 and YOLOv5,It is concluded that their average accuracy was 79.59%,87.37%,83.24%,91.18% and 92.80% respectively;Specifically,YOLOv5 reached 92.8%,39.4% and39.4FPS in terms of identification accuracy,recall and speed,respectively,Are higher than the other four target detection algorithms.According to the comparative analysis of the experimental results,YOLOv5 has a better detection effect in the identification of asphalt pavement cracks.(3)Based on the deficiency of the traditional detection methods for pavement diseases,Improving on the backbone network and the loss function of the YOLOv5,To propose an improved algorithmic model based on YOLOv5,And test on the datasets based on this,The training loss value of the improved YOLOv5 algorithm is always lower than that before the improvement,And keep it at about 0.015;The accuracy,recall,and average accuracy of the improved algorithm reached 95.5%,83.5%,95.1%,respectively,The detection speed is44.9FPS,By 5.1%,9.9%,2.3% and 5.5 FPS,respectively,This paper shows that the improved algorithm of this paper achieves good results.(4)According to the research situation of rural road asphalt pavement in Henan province,this paper analyzes various influencing factors in the maintenance decision process,combines different properties,establishes the "decision tree" model of maintenance countermeasures,and lists the corresponding maintenance measures.By assigning weight to each influencing factor,the mathematical model of multi-objective decision sorting method is established,the method of "weighted deviation sum of squares minimized" is adopted to prioritize each maintenance section,and the network 0-1 network-level decision model of rural road surface maintenance is established in the case of limited maintenance funds and maintenance personnel.This paper of rural highway asphalt pavement disease identification method and maintenance decision research make up for the shortage of traditional pavement disease detection method,for rural road asphalt pavement maintenance decision to provide an effective solution,at the same time for artificial intelligence technology of rural asphalt pavement identification and maintenance to provide certain reference and reference.