Research On Road Extraction Method For UAV Low Altitude High Resolution Image

With the wide application of low altitude remote sensing technology in the high resolution image acquisition,the low altitude remote sensing of UAV has now become a necessary technology for urban and rural road planning.UAV has many advantages,such as flexibility,speed of updating,and so on,which plays an important role in road extraction.The high resolution images obtained by the low altitude remote sensing of the urban and rural unmanned aerial vehicles are easily disturbed by the noise of other objects,such as trees,traffic signs,and architectural shadows,which are similar to the spectral or texture features of the road.Different degrees of difficulty.In view of the extraction of road between urban and rural areas,an automatic mixed characteristic road extraction method and process is proposed.The specific research contents include: combining the advantages of different color channels of the two color space models of RGB and HSI,a hybrid spectral feature extraction method is proposed.The traditional method of texture feature selection is more fixed and needs to be fixed.In artificial processing,this paper uses an improved class distance separability criterion to optimize the texture feature method.Using the above method,the spectral and texture features are extracted and the D-S evidence theory is used for feature fusion.The method of improving the likelihood function threshold determination based on evidence theory is used to determine whether the image is a road.The two value road map of the target image is extracted from the mixed feature,and the image of the two value map is processed by the method of regional marking combined with the geometric feature of the external rectangle,and the experimental evaluation is carried out by the Wiedemann quality evaluation method.The experimental data show that the method proposed in this paper has a good effect on road extraction in the urban andrural combination of low altitude and high resolution remote sensing images.The total rate of road extraction can reach more than 95%,and this method can be extended to other objects.