A Study On Spatial Invariant Features Based Fast 3D Reconstruction Technology Of Irregular Objects

Terrain and surface objects have many features, for example, they have complex contours and hard to be measured. When it comes to irregular surface objects, the 3D reconstruction has a lot of limitations, and measuring methods such as satellite position technologies, field measurement like total stations and low altitude remote sensing like unmanned aerial vehicles all have their own drawbacks. This dissertation mainly discusses a study on spatial invariant features and close-range photogrammetry based fast 3D reconstruction technology of irregular objects, and it provides a new technical method for volume calculation and visualization of irregular objects.In this dissertation, the research content mainly has the following four aspects:(1) A Delaunay construction algorithm oriented to 3D reconstruction of irregular objects has been designed. This algorithm combines topological relation, and it is based on the Delaunay triangulation network growth algorithm but adds some improvements:our new algorithm introduces new data members and function members, and finally realizes the automatic detection and visualization of the boundary of triangulation networks.(2) A new datum plane construction algorithm which uses the boundary information of triangulation networks for construction has been proposed taking advantage of the invariant features of some geometric quantities such as volume. Our algorithm uses the points on the boundary lines of the triangulation network, and calculate the plane equation using least square method. The plane we constructed could accurately represent the bottom contours of the triangulation network, and it provides a datum plane for projection of 3D points for volume calculation.(3) A self-adapted interpolation subdivision algorithm based on the detection of feature points has been designed. This method uses SIFT operator to extract the feature points between different images. When the number of feature points is less than the threshold value we set, we execute our subdivision algorithm in order to improve calculation accuracy. Experiment verified that our subdivision algorithm could effectively improve the precision of volume calculation.(4) A 3D reconstruction system for irregular objects has been developed based on our former accomplishments. This system is built on Windows platform, using Visual Studio 2010 and object-oriented technology as development tools, and finally realizes fast measurement and reconstruction of irregular objects with simple measuring tools.