Robot 3D shape recognition method based on multi-view information fusion

Abstract

The invention provides a robot 3D shape recognition method based on multi-view information fusion. The method combines the advantages of a full-view method and a single-view method, overcomes the disadvantages of the two methods, and comprises steps of: firstly, performing image similarity ordering by using an image similarity detection technique by means of the multi-view information of a 3D shape acquired by a robot in motion, so as to obtain a hierarchical depth feature by a convolutional neural network; finally, learning the visual feature with a certain time and spatial sequence by using a long short term memory model to obtain a highly abstract space-time feature. The method not only simulates the hierarchical learning mechanism of human beings, but also adds the space-time sequence learning mechanism for simulating the human learning, and realizes the high-precision classification and recognition of 3D shape by multi-view information fusion.

Description

technical field [0001] The present invention relates to the field of robot technology and computer vision. Specifically, it utilizes the multi-view visual information obtained by the visual sensor of the robot, and uses the hierarchical deep learning network, the time-space sequence deep learning network and the image similarity detection and sorting technology to realize the robot's alignment. Recognition and classification of 3D shapes. Background technique [0002] 3D shape recognition has always been a hotspot in the field of robotics and computer vision. Fast and efficient recognition of three-dimensional shapes is of great significance to real life. For example, robots or unmanned aerial vehicles can quickly retrieve and identify objects in the database through three-dimensional shape matching, and use them to find and determine targets or avoid obstacles and improve their own performance. Degree of intelligence; Public security and other fields use 3D matching technol...

AI Technical Summary

Problems solved by technology

[0005] In summary, in the 3D shape recognition method for robots, although the full viewing angle can obtain relatively high recognition and classification accuracy, it has relatively large limitations in practical application, and it is unlikely that the robot can obtain the full viewing angle of the 3D shape. Visual information can often only perform recognition tasks based on visual information from partial perspectives

In addition, although the calculation speed of single-view is very fast, after all, there is only one visual information of the view, and the amount of information is not sufficient, resulting in low classification and recognition accuracy, which cannot meet the actual needs

Images