Research On Automatic Detection Method Of Regular Buildings Based On Parallax Image

The basic theory of building research on images of Remote Sensing began in the1980 s.With the development of high resolution remote sensing technology,the rapid and automatic extraction of surface building information by using optical remote sensing technology is of great significance to urban planning,urban illegal building supervision and management and other fields.Since there are many categories of remote sensing images,each of which has its own advantages and disadvantages,it is a developing trend to integrate different categories of remote sensing images to extract buildings.Based on high-resolution aerial image data and parallax image data,this study combines the spectral characteristics of aerial image data with the accurate visual difference cloud information in the parallax image,integrates the advantages of the two,and realizes the automatic extraction of regular urban buildings according to the characteristics of buildings that are different from other ground objects.The main tasks of this study are as follows:(1)3D/2D data fusion based on per-pixel density matching parallax.Firstly,according to the results of aerial triangulation,the nuclear line images were corrected one by one.Then,the parallax maps of stereo pairs were obtained by pixel-by-pixel intensive matching.Finally,the 3D/2D data were fused by combining the 2D vector information obtained from field mapping to obtain the final parallax image data.(2)The octree algorithm was used to segment the visual threshold cloud data by voxelization,and case-based reasoning was used to classify the voxels.The depth information of the original image in the experimental area is recovered by preprocessing the parallax image.The visual threshold cloud information of the original image in the experimental area was extracted by voxel extraction based on octree point cloud segmentation algorithm.The attribute features and spatial features were constructed,the inference model was established,and the weight of the characteristic parameters was determined.After the inference,the voxels were classified and the building location information was extracted.(3)After obtaining the position information of regular buildings,it is returned to the original image to extract the contour of regular buildings at relevant positions.In this paper,two methods are used to extract the contour of regular buildings.First,building contour is extracted based on grid sequence method,and each unit is processed and analyzed to determine its contour segment.Then,contour distribution is formed according to the above results.Based on the region growth method,building contour extraction was carried out.Starting from the set of seed points,contour unit was found to complete the quantitative analysis of contour,and automatic extraction of regular building contour was realized.(4)On the basis of theoretical study,based on MATLAB R2014 a software platform,the experiment data of the experiment area,by means of building extraction result of precision analysis and the contrast between the different algorithms to analyze,verify the feasibility and reliability of the algorithm in this paper,and the extraction rules buildings show the advantages of integrity and accuracy.