Research On Road Network Generation Algorithms Based On Vehicle Trajectories And Remote Sensing Images

The development of the city is inseparable from the construction of the road network,and an accurate and complete road network map is the premise of urban traffic activities such as traffic navigation and route planning.The continuous upgrading and transformation of the city will inevitably have an impact on the road network in the city.However,the traditional road network map generation method based on manual mapping has the disadvantages of long cycle and high cost.Therefore,an efficient and economical road network generation and update method has great application value.With the popularization of mobile internet technologies and satellite positioning technologies,it has become more and more easy to acquire the trajectories of moving objects.Nowadays,most vehicles are equipped with positioning and navigation devices.These vehicles generate a large amount of trajectory data every day,these trajectory data reveal the shape and evolution of the road network.However,because the acquisition of vehicle trajectory data is difficult,the coverage of the generated road network is low.The maturity of aerospace technology makes it easy to acquire high-resolution remote sensing images.High-definition remote sensing images contain different levels of roads.However,it is difficult to accurately extract the road network from the high-definition remote sensing image,and the timeliness of high-definition remote sensing images is poor.This article combines the advantages of these two categories of data.,two efficient and high-precision road network extraction algorithms are proposed in this paper.Firstly,an incremental learning algorithm for vehicle trajectory and road network extraction is proposed.The algorithm first preprocesses the input vehicle trajectory data,and then records the position information and timing information in the vehicle trajectory through representative points and connecting segments respectively.Finally,the road network map is generated incrementally by combining Delaunay triangulation and Dijkstra's algorithm.It does not need the existing road network map as the basis and uses the information contained in the vehicle trajectory data to generate the road network map incrementally.The accuracy of the road network is constantly increasing with the input of trajectory data.Secondly,a method of road area recognition from high-definition remote sensing images using convolutional neural network is proposed.In this method,a convolutional neural network model with regional characteristics is trained by using the region with high accuracy selected from the road network extracted from the above-mentioned vehicle trajectories as the label set of samples and the remote sensing image of the corresponding region downloaded from the Google map as the sample set.Then the convolution neural network model is used to identify the high-definition remote sensing image of the corresponding area and extract the road area.Furthermore,the recognition result of convolution neural network can be combined with the road network generated by the above-mentioned vehicle trajectory.The results show that the synthesized road network can complement each other and cover a wider area.This paper presents an incremental road network generation algorithm based on vehicle trajectory and a convolutional neural network algorithm for road extraction based on high-definition remote sensing images.These two methods can work independently and efficiently to generate their own road network,but also can work together to combine the advantages of the two methods to get a better road network.