Incomplete point cloud completion method based on hidden space topological structure constraint

Abstract

The invention discloses an incomplete point cloud completion method based on hidden space topological structure constraints, and belongs to the technical field of computer vision. On the basis of a three-dimensional shape complementing technology of the single-view incomplete point cloud, complementing the single-view incomplete point cloud obtained by converting the depth map to obtain a complete three-dimensional point cloud shape of the object, and realizing rapid reconstruction of geometric information of the perceived object. S1, collecting three-dimensional space coordinates of a target object; S2, performing point cloud attitude normalization on the original incomplete point cloud input by the system; S3, predicting a corresponding complete key point skeleton based on the attitude normalized incomplete point cloud; S4, based on the predicted complete key point skeleton, recovering a dense complete point cloud corresponding to the predicted complete key point skeleton; and S5, carrying out optimization of robot sensing task based on the complete three-dimensional point cloud. According to the method, shape completion can be carried out on the real three-dimensional point cloud of any pose, the generalization and robustness are higher, the application range is wide, the speed is high, and the anti-noise capability is high.

Description

technical field [0001] The invention relates to a cloud completion method for residual defects based on hidden space topological structure constraints, and belongs to the technical field of computer vision. Background technique [0002] As an application complex of modern computer, automatic control, mechanical manufacturing and other technologies, robots have extremely high autonomous decision-making and execution capabilities, and can replace humans to complete many complex tasks. The perception link in robot technology is an important interface for robots to realize environmental interaction. The robot interacts with the environment through perception means, and obtains various environmental object information to assist the operator in making decisions. With the development of computer vision technology in recent years, robot perception technology based on computer vision has been widely used. Robots collect three-dimensional information of objects through visual sensors ...

AI Technical Summary

Problems solved by technology

However, in the actual information collection process, due to the influence of factors such as occlusion, environmental noise, and equipment errors, the 3D information obtained directly (the present invention uses 3D point cloud as the representation form of 3D information) is often incomplete, and there is information loss, so It is impossible to describe and characterize the complete geometric shape of the object, which brings certain difficulties to the perception understanding and action planning based on the complete 3D model of the object in subsequent tasks

Images