Grabbing attitude estimation method based on image instance segmentation and point cloud PCA algorithm

Abstract

The invention discloses a grabbing attitude estimation method based on image instance segmentation and a point cloud PCA algorithm, and the method comprises the following steps: S1, collecting the image data of a grabbed object, and obtaining an RGB image and a depth image; S2, segmenting the acquired data by adopting a point cloud segmentation algorithm based on a Mask R-CNN; S3, performing point cloud denoising filtering on the segmented data; S4, calculating a homogeneous coordinate matrix for the denoised and filtered data through a point cloud PCA algorithm, and obtaining a grabbing pose. The method has the advantages of being small in calculation amount, high in stability and insensitive to the environment, objects placed at any position in the space can be grabbed, and the grabbing accuracy is greatly improved.

Description

technical field [0001] The invention relates to the technical field of out-of-order sorting of mechanical arms, in particular to a grasping attitude estimation method based on image instance segmentation and point cloud PCA algorithm. Background technique [0002] The random sorting of robotic arms (including industrial robots, collaborative robots, etc.) can be applied to various industrial or civilian scenarios, such as automated production lines, smart pharmacies, etc. For disorderly sorting in industrial scenarios, 2D cameras are often used. First, a 2D image template of the object to be grasped is established, and the 2D image template is used for matching and positioning, and the coordinates are transferred from the camera coordinate system to the robot coordinate system to obtain the captured object. Pose, the robot arm moves to the grab pose to grab. The advantage of this method is that it is stable and reliable, and has high precision, but it also has two disadvant...

AI Technical Summary

Problems solved by technology

The advantage of this method is that it is stable and reliable, and has high precision, but it also has two disadvantages: the first is poor generalization, and a two-dimensional image template must be established in advance for each object to be grasped; the second is that the grasping posture is two-dimensional , it is impossible to grasp the three-dimensional objects placed arbitrarily in space

The advantage of this method is that it is relatively direct and has strong generalization, but it is not stable enough, the accuracy of capture is low, and the amount of calculation is large, so it is not easy to deploy in real-time systems

Images