Three-dimensional reconstruction method considering multi-stage matching propagation of weak texture area

Abstract

A three-dimensional reconstruction method considering multi-stage matching propagation of a weak texture area comprises the following steps of completing pixel-level matching transmission through an existing algorithm, while creating an initial three-dimensional model to complete spatial plane reconstruction of an area rich in texture; based on image segmentation and the three-dimensional model constraint constructed in a first stage, completing matching of a weak texture region and spatial plane reconstruction under an energy function minimization framework; under an energy function minimization framework, adopting a multi-constraint optimization method and the like to solve matching and reliable three-dimensional reconstruction of a texture-free area. According to the method, the problemthat in the prior art, the image matching and reconstruction effect is generally not ideal for areas with weak textures, areas without textures and areas with repeated textures is solved, and the phenomena of 'jolt 'of the water surface, 'cavities' of the wall body and the like in a three-dimensional reconstruction result are avoided. The influence of factors such as illumination change, perspective distortion, weak texture and repeated texture areas in a scene is overcome, and a guarantee is provided for a real effect and surveying and mapping level precision.

Description

technical field [0001] The invention relates to the field of analysis and measurement, in particular to a three-dimensional reconstruction method considering multi-stage matching propagation in weak texture areas. Background technique [0002] Three-dimensional reconstruction, that is, the process of recovering the geometric structure of the surface of the object from the two-dimensional image, has always been an important research content of digital photogrammetry and computer vision. Compared with the traditional upright photogrammetry, since 3D reconstruction needs to obtain side texture information, oblique photography technology emerges at the historic moment, which comprehensively perceives complex scenes in a large-scale, multi-angle, high-precision, and high-definition way. Through efficient data The data results generated by the collection equipment and professional data processing process can intuitively reflect the appearance, position, height and other attributes...

AI Technical Summary

Problems solved by technology

However, due to the complexity of the scene structure (especially for urban scenes), the influence of factors such as illumination changes, perspective distortion, weak texture and repeated texture areas in the scene, it is still difficult to effectively obtain complete 3D reconstruction results using oblique multi-view image information. is a very challenging subject

Among the existing related algorithms and relatively mature commercial software, the matching and 3D reconstruction of texture-rich images or image regions are relatively perfect, but for some weak texture, no texture and repeated texture regions, image matching and reconstruction The effect is usually not ideal, which leads to phenomena such as "bumps" on the water surface and "holes" in the walls in the subsequent 3D reconstruction results

Images