Research On The Detection Of Road Impact Factors Through Remote Sensing

The main factors that influencing the traffic conditions include:road conditions,vehicle performance,traffic conditions,environment,climate,etc.The "road conditions" refers to the road geometry.The "environment" refers to the landscapes of the road,natural conditions,road blocking,etc.Theses factors are more suitable for applying modern remote sensing technology to carry out investigation.Domestic GF images,which are good source for monitoring traffic factors,have many advantages such as high resolution,rich texture information and real time,etc.At present,there are many problems of extracting the information of the factors from remote sensing images,which influencing the traffic.The mian problems are:unintelligent and low degree of automation of the technology and methods,higher requirements for professional knowledge and greater influence on the traditional change detection methods such as image conditions,spectral difference and interference information,etc.In the view of the above problems,this paper applies the GF-1 images and studies the methods of extracting road centerline based on data mining and detecting traffic blocking points based on SIFT algorithm.Firstly,in the preprocess of decision tree classification,use data mining algorithm to generate decision tree and then extract the road and the results are post-processed based on ArcGIS to extract the road centerline.Then,evaluate the effectiveness of the method.Secondly,based on the second development of CAD-VBA platform,use mathematical statistics and least square method to calculate the curve radius parameters and the slope information of the road.Finally,detect and locate the traffic blocking points based on SIFT algorithm.The following main conclusions are drawn through the experimental research and analysis.(1)The C4.5 algorithm can complete the automatic and intelligent generation of the road extraction rules.The mean square error of road information extraction is similar to that of the object oriented method,which is respectively ±3.143m and ±2.904m.Using GF-1 images can extract the paved roads and unpaved roads integrated simultaneously based on C4.5 algorithm through the convert program which is written based on C#.The experimental results show that in the process of the northwest paved road extraction,through the histogram equalization and other brightness stretching operation,the decision rules of the road extraction are universal.(2)The algorithm of calculating road linear parameters based on CAD-VBA can calculate the curvature radius and slope parameters of the road network accurately and quickly.The relative error of the RTK measured data is 0.19%and the relative error of the simulation line is 0.14%.The results of expreiments show that this mthod can calculate the radius of the road with easrment curve,and the relative error is 0.36%.The accuracy of slope calculation is strongly correlated with the accuracy of road extraction and DEM resolution.The results show that it is approprate to choose 150m to sample points from road with DEM with 30m resolution.(3)The technical scheme for detecting traffic blocking points based on SIFT algorithm has a large amount of computation and the requirement to hardware is high.But the influences of this technology are significantly lower than that of the traditional change detection method,which has higher practical value.