Images Sequences-based Method For Fast 3D Reconstruction Of China Ancient Building

The method of 3D reconstruction based on images is a combination of Multiple view geometry and Computer vision algorithm and theory, it aims at recovering the 3D model of object and scene from a series of continuous images.Recently, the application of 3D reconstruction in military and civilian fields are becoming more and more, and one important application of those is study in reconstructing digital model of antient buildings. Old buildings is one important part of culture heritage for cohesion in China historic culture and wisdom. As a result of natural or man-made damage, the store and recovering of ancient buildings is urgent need for digital protecting and rendering. At present, it still has many difficulties remained. The paper is dedicated in discussing space 3D model reconstruction procedures and methods of old buildings from its images sequences by combining classic and novel algorithms in computer vision, The main study in the paper is that:1)A novel matching method using triangulation network based on SIFT (Scale Invariant Feature Transform) is proposed in settling the problem that the false match rate is high for complex scenes in ancient buildings. Triangulation network is formed by connecting good feature points, it express images scene’s structure and content, not only that network is adaptive for texture information but also it use the connection information among better local feature points to restrain matching search region. It is verified by experiments that the method is invariant for image zooming, translation, rotation, illumination variation, affine and ambiguity change to a certain degree and improve the matching accuracy and repetition rate.2) By combining computer vision’s classic and novel algorithm and studying the key technolges and algorithms in 3D reconstruction based on images, space structure information and part of 3D model is recovered from ancient building images sequences with the feature point matching method proposed, according to the experiment, triangle grid is formed correctly using the proposed method, the grid is adapt to the change of texture automatically,and the matching accuracy is improved efficiently.