Iterative 6D pose estimation method and device based on deep learning

Abstract

The invention relates to an iterative 6D pose estimation method and device based on deep learning, and the method comprises the steps: carrying out the coding of the output of a 3D-2D key point corresponding relation extraction network, and coding the output of the 3D-2D key point corresponding relation extraction network into an input format of a pose coarse estimation network; the method comprises the following steps that: firstly, object 6D postures are input into a pose coarse estimation network constructed by utilizing an MLP, a pooling layer and a full connection layer, so that the pose coarse estimation network is combined into an integral network, the object 6D postures can be directly output, an end-to-end 6D object pose estimation network is formed, and a loss function of the 6D object pose estimation network is a function capable of directly reflecting and resolving object 6D pose parameters; and optimizing 6D attitude parameters output by the 6D object pose estimation network by using an orthogonal iterative algorithm. According to the method, the problems of long time consumption and poor repeatability of calculating the target attitude based on the PnP algorithm when many abnormal values exist are solved, and the attitude estimation efficiency and the attitude estimation robustness and accuracy are improved.

Description

technical field [0001] The invention belongs to the field of object 6D pose estimation based on monocular vision, and in particular relates to an iterative 6D pose estimation method and device based on deep learning. Background technique [0002] Object 6D pose estimation refers to estimating the 6D pose of the object in the camera coordinate system, including rotation and translation, that is, obtaining the rotation and translation transformation from the object's own coordinate system to the camera coordinate system. Estimating the 6D pose of an object from a monocular RGB image is a fundamental problem in computer vision. It has many important applications such as robotic grasping, autonomous navigation, augmented reality, etc. Along with the development of depth cameras, many recent solutions are based on depth maps. However, depth cameras have limitations in frame rate, field of view, resolution, and depth range, and it is difficult to detect small, thin, transparent,...

AI Technical Summary

Problems solved by technology

The main problems of this type of method are: first, the loss function used to train the deep network focuses on finding the 2D key points of the projection of the 3D key points of the object in the image, and does not reflect the real purpose of pose estimation; secondly, solving the 6D pose of the object The PnP algorithm is not part of the network, which makes the entire network not end-to-end trainable; third, the RANSAC algorithm is very time-consuming, especially when there are a large number of outliers

Therefore, pose estimation that satisfies both accurate and fast robust requirements is a challenging problem

Images