RGB-D feature matching-based 3D point cloud registration and fusion method

Abstract

The invention discloses an RGB-D feature matching-based 3D point cloud registration and fusion method. The method mainly comprises the following steps of: projecting scene surface 3D point cloud dataof different viewpoints to RGB-D images to carry out processing, so as to realize dimensionality reduction calculation from three-dimensional point cloud to two-dimensional point cloud; extracting feature points of the RGB-D images, establishing feature descriptions of the RGB-D images, and matching the RGB-D images of the different viewpoints; solving a coordinate and a depth transformation matrix, and registering and fusing the RGB-D images; and converting the fused RGB-D images into 3D point cloud data. Through the method provided by the invention, calculation of three-dimensional feature extraction and matching can be simplified, and the efficiency of three-dimensional point cloud registration and fusion can be improved. The method provided by the invention can be applied to scene reconstruction of 3D large fields of view and 3D full views.

Description

technical field [0001] The invention relates to the fields of computer vision and digital image processing, in particular to a 3D point cloud splicing and fusion method based on RGB-D feature matching. Background technique [0002] Scene 3D reconstruction is an important research topic in the field of computer vision. In practical applications, the complete 3D point cloud data of the surface of the measured scene cannot be obtained at one time. The point cloud data can only be obtained through multi-viewpoint shooting or measurement, and then registered to a unified coordinate. Department. The commonly used method: relying on instrument registration to rotate and translate multiple point clouds into a unified coordinate system through the coordinate relationship between hardware, this method does not require common features between point clouds, and the registration speed is fast, but requires hardware Accurate calibration or the assistance of marker points to obtain the co...

AI Technical Summary

Problems solved by technology

[0002] Scene 3D reconstruction is an important research topic in the field of computer vision. In practical applications, the complete 3D point cloud data of the measured scene surface cannot be obtained at one time. The point cloud data can only be obtained through multi-viewpoint shooting or measurement, and then registered to a unified coordinate. Department

The commonly used method: relying on instrument registration to rotate and translate multiple point clouds into a unified coordinate system through the coordinate relationship between hardware, this method does not require common features between point clouds, and the registration speed is fast, but requires hardware Accurate calibration or the assistance of marker points to obtain the coordinate transformation relationship between the whole scene images; automatic registration is to eliminate the misalignment between the two clouds through a certain algorithm or statistical law. This method requires that the two point cloud data have common features , in some cases it is necessary to obtain the initial value, and the registration algorithm degenerates

Images