A point cloud self-registration method based on calibration sphere

Abstract

The invention discloses a point cloud self-registration method based on a calibration sphere. The method based on the calibration sphere is used for point cloud registration. The calibration sphere is used as the control point of the same name, and the Kinectv3 depth sensor is used to shoot the target object and the calibration sphere from multiple angles. The surface point cloud information, through point cloud processing, finally realizes the establishment of the target object model. Based on the comparison of the registration methods in the common overlapping area, the invention can perform large-angle point cloud registration, does not require high overlapping degree of two point clouds, is suitable for operation in complex and changeable environments, and proposes an automatic matching of the same name. The point method eliminates manual operations and makes the entire registration process more automated.

Description

technical field [0001] The invention belongs to the technical field of point cloud processing, in particular to a point cloud self-registration method based on a calibration sphere. Background technique [0002] At present, there are three methods of point cloud acquisition, namely binocular vision, time-of-flight, and structured light. (1) Binocular vision is the only passive measurement device in the three categories. It is essentially two identical cameras placed at different angles to shoot the same scene. Because of the different placement positions, the captured images are slightly different , and then by analyzing the difference between the two scenes, the depth information of each point in the image is finally deduced. The binocular vision device is cheap, but the accuracy is very low. Ideally, to get the spatial position of the pixel point, the two cameras need to be perfectly calibrated, but in fact, this calibration works in reality. It is impossible to implemen...

AI Technical Summary

Problems solved by technology

[0004] The data volume of the 3D point cloud scanned by the depth sensor is huge, and the whole registration process still takes a long time; there needs to be a large common overlapping area between the source point cloud and the target point cloud based on feature registration, so as to make The result is good; in the method of 3D point cloud registration based on the calibration object, after extracting the features of the calibration object, there is no method for automatically matching feature point pairs, and the overall process is not automatic enough

Images