Monocular 6D pose estimation method based on regular icosahedron

Abstract

The invention discloses a monocular 6D pose estimation method based on a regular icosahedron. According to the method, an ArUco code is combined with a regular icosahedron, then a monocular camera is used for collecting a regular icosahedron image marked by the ArUco code, the spatial three-dimensional coordinate of the central point of the ArUco code is calculated, the sparse point cloud of an inscribed sphere of the ArUco code is obtained, then multiple nonlinear fitting and an iterative nearest point algorithm are used for comparing the virtual regular icosahedron at the coordinate of the camera. And further the spatial displacement and the rotation angle of the current regular icosahedron can be calculated, and the pose of the measured object is calculated indirectly. The method overcomes the high requirements of an existing method for illumination, environment and equipment, the detection is rapid, and the stability is improved; according to the method, a large amount of energy does not need to be put into template acquisition or model training in the early stage, the pose estimation cost is remarkably reduced, the pose estimation precision is ensured, the universality is enhanced, and pose estimation of large-amplitude movement of an object can be carried out.

Description

technical field [0001] The invention belongs to the field of image recognition pose detection, in particular to a monocular 6D pose estimation method based on a regular icosahedron. Background technique [0002] The pose estimation of the target plays an important role in both production and life, and is the prerequisite for flexible and effective robot movement. At present, the industrial automation industry is developing rapidly, and the requirements for real-time detection and feedback of the state of the robot are also increasing, and the estimation of the target pose based on vision is of great significance to the improvement of the performance of the robot. The monocular vision system uses only one camera, has a simple structure and low cost, and is widely used. At present, the main methods of target pose estimation are: [0003] Document "LEPETIT V,PILET J,FUA P.Point matching as a classification problem for fast and robust object pose estimation[C] / / Proceedings of ...

AI Technical Summary

Problems solved by technology

In the document "REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: Unified, real-time object detection [C] / / Proceedings of the IEEE conference on computer vision and pattern recognition.2016:779-788." The YOLO network restores its pose through the two-dimensional frame of the target in the image, but the method based on deep learning requires a large number of sample resources to construct a large amount of sample data and mark the real pose, and requires a large amount of computing resources for network training. It is difficult to control the cost, and most of them are targeted development, which cannot meet the versatility

Images