Anti-occlusion object pose estimation method based on deep neural network

Abstract

The invention provides an anti-occlusion object pose estimation method based on a deep neural network. The anti-occlusion object pose estimation method comprises the following steps: constructing a training set picture database and a test set picture database with labels by utilizing 3D modeling software; constructing a deep neural network, wherein the neural network comprises four sub-branch networks, and each sub-branch is of an independent convolutional neural network structure; constructing a network prediction output processing algorithm: four sub-branch networks output 6-dimensional prediction values at the last dense connection layer, wherein the prediction values represent pose information of the to-be-estimated object; wherein due to the existence of shielding, the result of a certain branch has an error, abnormal values exist in the output of the four branches, five algorithms are constructed to optimize the abnormal values, and the shielding interference resistance is improved; training a deep neural network model: using samples in the training set to complete training of the deep neural network; and testing the deep neural network model by using different shielding proportion test sets.

Description

technical field [0001] The invention belongs to the field of object pose estimation, and relates to a method for estimating an object pose with strong anti-interference performance using a deep neural network. Background technique [0002] The object pose covers all the spatial information of the object, including the position information and attitude information of the object. In many fields such as modern industrial production and life, the pose information of objects is of great significance and plays a pivotal role. Accurately estimating the pose information of objects is the basis of many current industrial applications. For example, in the field of robotics, accurate acquisition of target position and attitude information is the main task of robot vision and the basis for other subsequent operations such as grasping. In the field of IoT autonomous driving, accurate estimation of obstacle poses is the premise and guarantee for safe driving. Therefore, it is of great ...

AI Technical Summary

Problems solved by technology

The other is that the target is blocked by interference. This situation is more common and common in industrial applications. However, at present, only increasing the number of samples and diversity can be used to overcome the interference and reduce the impact. There is no effective solution.

Images