Fast object three-dimensional pose estimation method based on RGBD camera

Abstract

The invention discloses a fast object three-dimensional pose estimation method based on an RGBD camera. The estimation method is realized through the following modules: an image and spatial feature extraction and fusion module, a three-dimensional key point prediction module and a pose resolving module with a microseparable group elimination mechanism. The method is mainly characterized in that a point-by-point feature fusion mechanism is utilized, image features and spatial features of an object are utilized at the same time, the richness of the extracted features is increased, it is guaranteed that the features conform to the three-dimensional structure of the object, and the key point estimation precision is improved; a confidence coefficient weighted key point estimation mechanism is utilized, in a process without iterative loop, the influence of inaccurate estimation on the overall estimator is inhibited, the key point estimation precision is improved, and the time efficiency is ensured; by means of a differentiable outlier elimination mechanism, accurate object three-dimensional pose estimation based on the RGBD camera is achieved in a process which does not need iterative loop, and the precision of pose estimation is guaranteed.

Description

technical field [0001] The invention relates to the field of three-dimensional pose estimation, in particular to a fast object three-dimensional pose estimation method based on an RGBD camera. Background technique [0002] The 3D pose estimation of an object refers to estimating the relative pose relationship between the intrinsic coordinate system of the object and the camera coordinate system, including three-degree-of-freedom rotation and three-degree-of-freedom translation. The 3D pose estimation of objects has a wide range of applications in robot grasping, virtual reality, autonomous driving and other fields. According to the different sensors used, it can be divided into pose estimation methods based on RGB cameras and pose estimation methods based on RGBD cameras. Among them, the pose estimation method based on the RGB camera only uses the RGB camera as the sensor, and then extracts the object features from the image to estimate the pose. The equipment is simple, bu...

AI Technical Summary

Problems solved by technology

This type of method can obtain higher estimation accuracy than the traditional method, and has strong applicability, but the final optimization link takes more time and the real-time performance is relatively poor.

Images