A method of intelligently grasping image generation based on conditional generative adversarial network

Abstract

The invention discloses an intelligent capture image generation method based on a conditional generative confrontation network. The method first constructs a virtual and real capture environment and a conditional generative-confrontational neural network, and then uses the existing capture data sets to judge the capture quality respectively. device, image quality discriminator D PQ and generator G for cyclic iteration training. Finally, a noise-specific depth image is generated by the trained generator. The invention combines the high-precision mechanical structure of the robot with the high robustness of deep learning, and realizes intelligent and reliable grasping behavior for the robot when no specific task is given or the shape of the object to be sorted is relatively complex and the environment is relatively changeable. Provide a data base.

Description

technical field [0001] The invention belongs to the fields of intelligent manufacturing and machine learning, and in particular relates to an intelligent capture image generation method based on a conditional generative confrontation network. Background technique [0002] With the development of Industry 3.0, the initial automated robots undertake repetitive, boring and low-intelligence labor, liberating human beings. The robotic arm is one of the most common robots in the industry. At present, the robotic arm is widely used in industrial environments and even in home hospitals. Grabbing and moving objects is one of the most important tasks of the robotic arm. The advantage of the robotic arm is that it can quickly complete a given task with high precision. When the position, shape and posture of the object are fixed, the reasonable setting of the robotic arm's action can efficiently complete the grasping. However, with the trend and development of Industry 4.0, the above m...

AI Technical Summary

Problems solved by technology

However, with the trend and development of Industry 4.0, the above methods have revealed the following problems: robots are not only required to be able to perform repetitive tasks, but are also expected to be able to complete complex tasks to a certain extent and have the ability to respond to environmental changes

However, this calibration process takes a lot of time, and the grasping model is limited, resulting in poor generalization ability of the data set.

Images