A real-time indoor 3D reconstruction method based on low-cost RGB-D sensor

Abstract

The invention relates to a method based on a low-cost RGB-D sensor. Firstly, a series of disordered lists of panel elements are constructed to represent the reconstructed 3D scene. According to the time threshold, the reconstruction model is divided into an active model and an inactive model. The RGB-D sensor obtains the depth image and color image of the indoor scene, and obtains the sensor poseestimation of the latest frame through the joint optimization of geometric tracking and photometric tracking from the frame to the model, thus completing the preliminary three-dimensional reconstruction of the indoor scene; a deformation map is constructed for each frame of point cloud data, the deformation map is optimized to improve the registration accuracy of single frame model surface, and further improve the overall three-dimensional reconstruction accuracy; Finally, when the sensor re-accesses the reconstructed region, the active and inactive models are strongly aligned to complete thelocal loop detection, and the complete, accurate and high-precision indoor scene three-dimensional model is finally obtained.

Description

technical field [0001] The invention relates to the technical field of real-time three-dimensional reconstruction of indoor scenes, in particular to a real-time indoor three-dimensional reconstruction method based on a low-cost RGB-D sensor. Background technique [0002] 3D reconstruction technology has always been a hot topic in the field of computer graphics and computer vision. Early 3D reconstruction technology usually used 2D images as input to reconstruct the 3D model in the scene. However, limited by the lack of rich input data information, The reconstructed 3D models are usually incomplete, less accurate and less realistic. Subsequently, ranging sensors such as lasers and radars or structured light technologies were used to realize 3D reconstruction of scenes, but because these devices are expensive and inconvenient to use, their applications are limited. With the emergence of various consumer-grade RGB-D sensors, 3D scanning and reconstruction technologies based on...

AI Technical Summary

Problems solved by technology

[0002] 3D reconstruction technology has always been a hot topic in the field of computer graphics and computer vision. Early 3D reconstruction technology usually used 2D images as input to reconstruct the 3D model in the scene. However, limited by the lack of rich input data information, The reconstructed 3D models are usually incomplete, less accurate and less realistic

[0004] In addition, the depth map contains noise and can contain holes caused by occlusion and object surface characteristics, which is a challenge for high-quality reconstruction and is also the problem to be solved by the present invention

Images