Real-time three-dimensional reconstruction method for object grabbed by mechanical arm

Abstract

The invention relates to a real-time three-dimensional reconstruction method for an object grabbed by a mechanical arm, and the method comprises the following steps: (1) obtaining scene depth information and a color image through an RGB-D depth camera, carrying out the real-time background segmentation based on the depth information, removing a background from a reconstruction scene, and only retaining a dynamic object; (2) for the dynamic object retained in the step (1), obtaining a three-dimensional point cloud model of the dynamic object through real-time pose calculation of the dynamic object; and (3) for the three-dimensional point cloud model of the dynamic object obtained in the step (2), removing a mechanical claw in the three-dimensional point cloud model of the dynamic object to obtain a three-dimensional point cloud model of the object grabbed by the mechanical arm. Complete segmentation between the mechanical claw and the target model can be realized, the precision of the three-dimensional reconstruction model is improved under the influence of the mechanical claw, the segmentation accuracy of the model is improved, and the three-dimensional reconstruction speed is improved.

Description

technical field [0001] The invention relates to the field of real-time three-dimensional reconstruction of dynamic objects, in particular to a method for real-time three-dimensional reconstruction of objects grasped by a mechanical arm. Background technique [0002] Three-dimensional reconstruction refers to the process of establishing a suitable computer model for three-dimensional objects and performing digital processing. It is the most direct way for computers or intelligent robots to perceive the environment. It is widely used in all walks of life. How to obtain accurate models quickly and effectively, It has become an important research topic in both academic and industrial fields. [0003] With the continuous development of the field of 3D reconstruction research, the establishment of 3D models has gradually developed from the traditional contact method to a variety of non-contact methods represented by visual 3D reconstruction. Visual 3D reconstruction has the advant...

AI Technical Summary

Problems solved by technology

Among them, the dense 3D reconstruction algorithm based on multi-view has a long running time and cannot meet the needs of real-time 3D reconstruction. Therefore, the reconstruction method of visual SLAM capable of real-time 3D reconstruction has become the focus of research.

[0004] But at the same time, there are various constraints in the vision-based 3D reconstruction technology. First of all, at the principle level of the reconstruction project, the "static world assumption" is the common dependence of the current vision-based 3D reconstruction methods. The assumption stipulates that modeling All objects in the scene are absolutely static, but such a scene is difficult to achieve in practice, and the noise generated by other objects in the scene has a serious impact on the reconstruction results, making it difficult to popularize and apply 3D reconstruction technology in ordinary scenes; secondly, the construction of 3D models It is necessary to calculate the change of camera position and attitude in real time, which consumes a lot of computing resources; finally, when the reconstructed object and other objects in the scene occlude each other, and it is difficult for the model to obtain accurate prior information, how to accurately Segmenting the target object accurately is the current difficulty

Images