Research On Road Network Information Extraction And Update Based On Trajectory Data

With the continuous expansion of the city scale,the urban road network is in a rapid change,and timely access to accurate road network information becomes particularly important.Compared with traditional methods,obtaining road network information from vehicle track and street view image data has the advantages of low cost and high potential,but there are also problems such as uneven data distribution,low accuracy,high noise and how to obtain traffic signs from street view image.This paper focuses on the track and street view image data,proposes a method based on the track data to extract the road center line,road intersection,road boundary and road network update,and a method based on street view image to obtain the semantic information of traffic signs.The main research work of this paper is as follows:(1)The method of obtaining the road center line based on the track data was studied.The track raster image was obtained by rasterizing the track data.On this basis,the track image was processed by mathematical morphology and morphological skeleton extraction algorithm to obtain the road center line.Homomorphic filtering is introduced to improve the morphological skeleton extraction algorithm.Experiments show that the optimized algorithm has good denoising effect.(2)The method of extracting road intersections is studied.Based on the road centerline,this paper uses the pixel neighborhood intersection pixel discrimination method to construct the 8 neighborhood pixel frames to identify the number of pixels.Therefore,the road intersection is extracted.The experimental results show that the accuracy of the method is 6.06%in the extraction of road intersections.(3)The method of extracting road boundaries is studied.The density change index and the side length index are used as constraints to construct the Delaunay triangulation,and the road boundary discriminant model is established.Finally,the Canny edge detection operator is used to detect the road boundary.In the threshold(0,1),it has better noise resistance and better edge retention.(4)The method of road network update is studied,and the image difference method and the road network update method based on wavelet transform and SIFT are adopted.In this paper,the SIFT algorithm is optimized by RANSAC algorithm to improve the image difference method.The experimental results show that the noise side of the road network has better rejection effect.Finally,the road network is updated by wavelet transform and SIFT.(5)The full convolutional neural network is used for deep learning to realize semantic segmentation,and the acquisition of traffic sign information is preliminarily explored.