Extraction Of Road Intersection Information Based On Low Frequency Trajectories

With the rapid development of wireless transmission technology and spatial positioning technology,as well as the improvement of software and hardware computing capabilities,vehicle trajectory acquisition is becoming easier.Vehicle GPS trajectories have the advantages of low cost,fast update and large scale.In addition,they contain other attribute information,from which a large amount of digital information about the road network can be mined.Therefore,more and more researchers have begun to pay attention to road information based on trajectory mining.At present,many vehicles are equipped with GPS equipment,especially taxis that shuttle through urban streets every day.However the quality of taxi trajectory data is not high because of the cost control,which has some shortcomings such as low sampling frequency,low positioning accuracy,many noise points,and uneven data distribution.There are many challenges when the taxi trajectory data used to extract road information.Aiming at the above-mentioned shortcomings of low-frequency trajectories and combining the advantages of clustering calculation and image processing,this paper proposed an integrated method,which takes the location information extraction of road intersections as the priority and on the basis of intersection location recognizes turn information.The main research content of this article has the following two aspects:(1)Location extraction of road intersection based on fusion of vector data and raster data: This paper extracts the location information of road intersection in vector space and raster space respectively.In the vector space,the density peak clustering algorithm is used to identify intersections,taking into account the dense and sparse problems of different road sections,to overcome the shortcomings of uneven data distribution.In raster space,mathematical morphology is used to process trajectory raster images with different resolutions to ensure the integrity of the results.This paper has designed a vector grid fusion rule to detect intersections,which performs fusion processing on multiple results,and judges the authenticity of intersections,removes false intersections to improve the accuracy of results.(2)Recognition of intersection turning information taking trajectory path into account: This paper recognize intersection turning information through analysis and calculation of trajectory path.This paper proposes to use Delaunay triangulation to detect the area radiated by intersections,and extract the trajectory subsequence in the radiation area as the trajectory path.Based on the principle of similarity measurement and clustering by direction,the direction change and the turning rules of the trajectory path is calculated and extracted.Finally,the associated intersection is identified according to the attribute information of the trajectory subsequence,and the turning relationship of the intersections is derived.The method in this paper is simple in calculation,fast in generating results,and makes full use of the trajectory path to overcome the limitations of low-frequency data and ensure the accuracy of the results.In order to verify the effectiveness of the research method,this paper takes the 10-day taxi trajectory data in Wuhan as the experimental data,and selects the Hankou area of Wuhan,which has a representative road network shape structure,as the experimental area.Compared with the results of other methods,the experimental results of this method are better in visual inspection and precision assessment.The method in this paper can be applied to low-frequency trajectory data,and it also has a good effect on sparse data areas.The results of road intersections location and turning information have high completeness and accuracy.Therefore,the method proposed in this paper provides a feasible supplementary scheme for enriching and updating the data of the navigable road network,and has certain application value in the intelligent transportation system.