A Dense Matching Algorithm Of Muiti-view Image By The Object-side Triangle Constraint

Multi-view matching technology is a core issue of photogrammetry and remote sensing direction which the essence is the process of identifying the same features from multi images,is widely used in various fields of computer vision, target reconstruction and 3D reconstruction. For the widespread issue of building block in urban image,this paper proposes the triangulation-based multi-view image constraints dense matching method, based on the status of the multi-view image matching technique,which takes advantage of multi-view image matching. The main contents of this paper are as follows:(1) Initial Match:Analysis of existing matching algorithms, introduced three types of constraint conditions and typical algorithms, according to the comparison of results in the VLL, Geometric constrained cross-correlation method and the mobile elevation plane constraint method experiments,this paper determine to use the results of mobile elevation plane constraint method to match,and to utilize result aloue as the initial dense matching data.(2) Occlusion detection:On the basis of analysing the existing typical of occlusion detection algorithm, and proposed a triangulation elevation of occlusion detection algorithm. In order to the elevation values "ground element" on grid plane based on feature point matching contains "a known point",irregular triangulated network is be built on the basis of feature point matching results to estimate elevation value in the triangulation,and constraint the elevation to detect occlusion. And verificating its feasibility through experiment.(3) Dense multi-view images match:Propose a method of triangulation constraints dense multi-view image matching, starting from the object side, the object side of plane is divided into the ground small elements,according to the visible situation after occlusion detection; By constructing irregular triang-ulated network to estimate to matched initial elevation of "ground element" to be matched, and the dense matching of this experiment according to the plumb line constraint.