Research On Construction Of High-precision Road Network In Open-pit Mine Based Oblique Photography

With the rapid development of unmanned and intelligent dispatching technologies in open-pit mines,unmanned vehicles in open-pit mines urgently need a mining road network model that is more in line with the actual conditions of open-pit mines in the daily production process.As an important part of open-pit mine transportation management,open-pit mine road network data is a solid foundation for building future smart mines.Especially in recent years,the domestic unmanned technology has gradually begun to be popularized and applied,and how to efficiently extract the road network model of the open-pit mine has become the key to the unmanned scheduling operation of the open-pit mine.This paper uses UAV oblique photography technology to collect a large number of road images in open pit mines,and extracts road information in mining areas through an improved deep learning model.Finally,based on road information,a highprecision road network for open pit mines that can be used for unmanned driving is constructed..The specific research work is as follows:(1)The road image data of the mining area is obtained by collecting the roads in the open-pit mining area through the UAV oblique photography technology,and the roads in the mining area are classified according to the imaging characteristics of the mining area.The road images in the mining area are marked as road areas and non-road areas,which are used for subsequent training samples of the deep learning model.(2)Roads in open-pit mines are unstructured roads,and their internal geometric information is more complex.Aiming at the problem of high plane shape and connectivity dimension after road analysis,this paper proposes MD-LinkNeSt(Multi D-LinkNeSt)which is suitable for extracting road information of open pit mines by improving the network structure and optimization function of D-Link Net network.network model.(3)By combining the road information extracted from the MD-LinkNeSt network with the image acquisition properties of the original oblique photographic image,an oblique photographic image of an open pit mine with road information that can be directly modeled is generated.In this paper,Agisoft Photo Scan software is used to complete the preliminary extraction of the road network model of the open pit mine,and repair the defects in the construction of the road network.On the basis of this model,a clustering algorithm is used to aggregate the open-pit mine road centerline from the orthomap of the open-pit mine road network to generate the open-pit mine road network centerline model.(4)In order to quantitatively analyze the accuracy of the open-pit mine road network model and the centerline model,this paper analyzes the accuracy,recall rate,coverage rate,and error rate of the two models.Finally,this paper uses Unity software to simulate the driving and meeting of unmanned vehicles in open-pit mines on the road network.The results show that the road network constructed by this method can meet the navigation accuracy requirements of unmanned vehicles in open-pit mines.This paper hopes to find a new method for efficient construction of the road network model of the open pit mine by extracting the road information of the open pit mine from the images collected by the UAV oblique photography technology and constructing the road network model of the open pit mine on this information.This paper uses UAV oblique photography technology and deep learning method to provide relevant scholars with a new research idea for the construction of road network model in open pit mines.The research advances the development of unmanned technology in open-pit mines...