Research On Intelligent Identification Of Bridge Traffic Information And Load Characteristics Based On Aerial Images

Vehicle load is one of the main loads of bridges.With the rapid development of my country’s economy and transportation,heavy loads have increased significantly,and higher requirements have been placed on the carrying capacity of the bridges in service.In the current design standards,there may be differences in regional traffic,especially in heavy vehicle loads.Some suburban bridges may underestimate the load.This paper uses drone aerial photography to collect traffic images of Xiangtan First Bridge,Xiangtan Third Bridge and Xiangtan Fourth Bridge,and establish a YOLO-V3 neural network model to identify the number of vehicles,models,distances,speeds and distances of traffic flows on each bridge.information.Combining the statistical data of the parameters of each vehicle type,the statistical value of the vehicle load concentration of each bridge is obtained,and the causes and effects of the load of each bridge are analyzed.The specific work content is as follows:(1)In order to obtain traffic flow data of three bridges in Xiangtan City,the three bridges in Xiangtan City were subjected to UAV aerial photography and on-site manual collection methods to conduct periodic traffic surveys on each bridge,and obtained characteristic data such as the main models and the number of vehicles on each bridge.The collected data was compared and verified with manual counting data,which verified the feasibility of the UAV collection method.The basic data of the traffic flow of each bridge was counted,and a sample data set of the characteristic information of each bridge was established.(2)In order to realize the recognition of aerial vehicle images,based on the principle of image analysis,a YOLO-V3 neural network model is established,and data such as the number of vehicles,vehicle types,distance between vehicles and vehicle speed are obtained by the recognition.Classify and solve problems such as flying height,model function,and vehicle classification.For the shadow problem in the image,the RGB three-color principle of the image is used to extract and eliminate the shadow part.The accuracy rate of YOLO-V3 model detection is further improved,and the detection rate basically reaches 89%.(3)Based on the actual measurement data of the traffic survey on each bridge,the mainstream vehicles on the first and third bridges are obtained as Type 1 and Type 3 vehicles.The composition of the traffic flow on the fourth bridge is complex,ranging from Type 1 to Type 7 cars.The distance between each lane and the lane load are calculated.And take the95% standard value.Refer to the current "Urban Bridge Load Standards" and compare and analyze the vehicle loads of each bridge.It is concluded that the lane load of the middle lane of the first bridge and the third bridge during the intensive operation period is greater than the specified value,and the continuous heavy vehicle of the fourth bridge has an impact on the safety of the bridge Larger.