Research On Safety Supervision Method Of Traffic Engineering Construction Site Based On Aerial Image

In recent years,China is in the period of large-scale infrastructure construction,and construction starts at the same time across the country.However,there are insufficient personnel,poor supervision,untimely problem handling,and difficulty in collecting evidence in the safety supervision of traffic engineering construction sites.The problem has caused more safety hazards on the construction site,frequent accidents,resulting in serious life and economic losses,accompanied by huge social adverse effects.With the development of social economy and technology,the technology of Unmanned aerial vehicle(UAV)is becoming more and more mature.Its safety,flexibility,high cost performance and other advantages have greatly facilitated the inspection work of various industries,and gradually applied to the safety of traffic engineering construction sites.In the supervision work,modern,automated and efficient safety supervision methods are provided for the construction site.This paper makes full use of the characteristics of the flexible maneuvering of the UAV,and combines the diversity characteristics of the traffic engineering construction work surface to collect all-round high-definition image data,realizes the three-dimensional and two-dimensional visualization of the construction site at each stage,and effectively supervises the actual situation of the construction site;Based on the deep learning technology,the automatic identification and classification of dangerous sources in the construction work surface can save a lot of manpower and material resources,greatly improve the processing efficiency of aerial image,and the aerial data can be retained,which provides an effective basis for subsequent inspection.The main research contents include:First,according to the characteristics of the traffic engineering construction site,the omni-directional images of the construction site are collected by the UAV.Plan the route of UAV,and use tilt camera to collect omni-directional image information of target for three-dimensional buildings.Collecting orthographic image information for plane construction.The three-dimensional reconstruction of the construction site is realized based on the inclined image,which can display the actual situation of the construction site.Then,image registration is carried out based on the position and attitude information of orthophoto image,and feature points are extracted from small target area of high-resolution image to modify homography matrix.Through the best seamline,adjacent images are stitched together to form a seamless high-resolution image with a large field of view.Finally,the two-dimensional visualization of the construction work surface can be realized to effectively supervise the construction progress,strengthen the site management,retain the site data and do follow-up inspection.Finally,based on the high-resolution panoramic mosaic image,the SSD convolutional neural network(CNN)algorithm is studied to automatically identify the dangerous source targets in the aerial image.Determine whether there are dangerous operations on the construction site and whether the installation distance of the antiglare plate of the expressway meets the requirements.Finally,the intelligent supervision effect is achieved,and the efficiency and accuracy of safety supervision on the construction site are greatly accelerated.