Study On Data-Model Driven Interior 3D Modeling

3D modeling is an important basic step in virtual reality technology.How to quickly and accurately construct an object’s three-dimensional model is related to the quality and precision of three-dimensional reconstruction.Aiming at the shortcomings of existing model-driven 3D modeling and data-driven 3D modeling,a data-model driven 3D modeling method is proposed for 3D reconstruction of small regular rectangles in indoor scenes.The research methods of this paper are as follows:1)Based on the spatial Euclidean distance algorithm of Kinect v2 equipment,the initial model is obtained by calculating the spatial vertical relationship of the model.2)Feature line segment extraction based on LSD algorithm;3)Iterative calculation of the three-dimensional space coordinate parameters of the eight vertices of the initial three-dimensional model based on generalized point photogrammetry theory,and at the same time to ensure the rectangular model line segment in the iterative process The spatial relationship between the two verticals does not change.The iterative calculation of the observation equation with the constraints is introduced,and the accurate three-dimensional coordinates of the model are finally calculated.4)Project the accurate model obtained into the line segment extraction result to realize the 3D modeling experiment.The experimental results show that:1)Kinect v2 sensor can obtain RGB-D data at low cost and convenient to construct the three-dimensional coordinates of the object obtained by color point cloud.The initial model acquisition experiment shows that the initial value of the model based on the data is automated and accurate.Both degree and work efficiency have advantages,and are also applicable to image data with low resolution and inconspicuous edge features;2)efficiency and accuracy in coordinate parameter iteration compared with classical generalized point photogrammetric 3D modeling method It has certain advantages,which is beneficial to improve the accuracy of 3D model and image registration.In summary,the data-model-driven 3D modeling method combines data-driven 3D modeling and model-driven 3D modeling methods with small data volume,high degree of automation,easy access to spatial 3D information,and better assurance of modeling accuracy.The amount of data is small,and the edge features of the object have obvious advantages.It can achieve 3D reconstruction with better effect and higher precision.