Three-dimensional reconstruction method and device of space plane relation constraint

Abstract

The invention discloses a three-dimensional reconstruction method and device based on space plane relation constraint, and the method comprises the steps: three common local and global scale space plane relations in indoor and urban scenes are introduced, and three support surface element sets are extracted for each reconstruction surface element; a supporting surface element is used as a carrier, and a space relation between the surface elements is used as an additional constraint to be introduced into a surface element generation stage, the negative effects of depth map noise and positioning errors are inhibited, the robustness, consistency and accuracy of a final three-dimensional reconstruction model are improved, the newly introduced part is small in calculated amount, the real-time performance and expandability of an existing scheme can be kept, and different scales of scenes can be adapted.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a three-dimensional reconstruction method and device constrained by spatial plane relations. Background technique [0002] The images collected by living electronic devices such as mobile phones are mainly two-dimensional color images, which are actually projections of three-dimensional imaging objects on a two-dimensional plane. Objects lack one dimension — depth. [0003] In order to make the imaging of objects in the three-dimensional space of the real world more three-dimensional and vivid, a three-dimensional reconstruction technology has emerged, which can construct a three-dimensional model of the imaging object, including three-dimensional position information (the dimension of depth is more than that of two-dimensional images), and The real world is the same. Therefore, the three-dimensional image obtained by three-dimensional reconstruction can be presented w...

AI Technical Summary

Problems solved by technology

[0011] 5. The requirements for hardware computing resources are low, and the computational complexity of 3D reconstruction is generally high. Some algorithms need to rely on multiple high-performance image computing units (GPUs), which hinders their practical application

[0014] However, the existing technical solutions only obtain the final reconstruction result of the surface element based on the fusion of information limited to different moments in the area of ​​the surface element itself, ignoring the relationship between the surface elements, resulting in too much freedom in the reconstruction of the surface element, and the reconstruction Surfs are susceptible to input depth image noise and camera positioning errors from the positioning system (in practical applications, depth image noise and positioning errors must exist), and the position and orientation of the reconstructed surfels may deviate significantly from the real value , and there may be inconsistencies between adjacent facets (such as the inconsistency of adjacent facets on the same plane)

Resulting in less robustness, consistency and accuracy of the final reconstructed 3D model

Images