Six-degree-of-freedom pose estimation algorithm based on bounding box outer key point positioning

Abstract

The invention discloses a six-degree-of-freedom pose estimation algorithm based on bounding box outer key point positioning, and the algorithm comprises the steps: inputting a three-dimensional modeland an RGB image, obtaining the size information of a target in the image in three dimensions, and determining the coordinates of eight vertexes of a target three-dimensional bounding box; detecting the RGB image through a convolutional neural network to obtain five feature maps with different scales, a target category of each position and a two-dimensional bounding box prediction result; carryingout non-maximum suppression operation to obtain the category of the target instance in the redundancy-removed image and a two-dimensional bounding box prediction result; positioning the key points inthe image to obtain positions and positioning confidence coefficients of the eight key points; and calculating the six-degree-of-freedom pose of the target relative to the camera by using an EPnP algorithm. The method can effectively improve the positioning precision of the key points outside the bounding box, and improves the pose estimation precision while guaranteeing the real-time processingcapability.

Description

technical field [0001] The invention belongs to the technical field of digital image processing, and more particularly relates to a six-degree-of-freedom pose estimation algorithm based on the positioning of key points outside the bounding box. Background technique [0002] Relative pose estimation with six degrees of freedom is a classic problem in the field of computer vision, but it still attracts the attention of a large number of researchers. Effectively obtaining the relative position and orientation of the object of interest is an important basis for realizing a variety of high-level vision tasks (such as augmented reality, autonomous driving, and robotics). Although the 6-DOF pose estimation method based on RGB-D data sources can achieve high accuracy, the pose estimation method based on RGB images has better efficiency and usability, so it has become a current research hotspot. Traditional 6-DOF pose estimation algorithms based on RGB images are often only suitable...

AI Technical Summary

Problems solved by technology

[0004] The prior art YOLO6D algorithm is based on the offset of the key points in the upper left corner of the grid. For key points in different positions, the length of the offset varies greatly, which is not conducive to the accurate prediction of the convolutional neural network; and most of the three-dimensional objects The vertices of the bounding box are located outside the grid, and the YOLO6D algorithm has limited positioning accuracy for the key points outside the grid; the positioning accuracy of different key points is different, and the YOLO6D algorithm does not consider the relationship between the key points when using the EPnP algorithm to calculate the pose. Due to the difference, the pose accuracy may be reduced by the influence of a few key points with large positioning errors

Images