Author(s)

Yifei Liu

Corresponding Author

Yifei Liu

Abstract

The purpose of this article is to research and develop a visual recognition and grasping system for industrial robots based on deep learning, so as to improve the recognition accuracy and grasping success rate of industrial robots in complex working environments. Aiming at the limitations of traditional methods in dealing with complex background and illumination changes, this article proposes an innovative solution combining deep learning, machine vision and industrial robot technology. By constructing an efficient deep learning model and optimizing the network structure and hyperparameters, the high-precision recognition of the target object is realized. At the same time, this article introduces the intelligent grasping strategy planning module, and uses reinforcement learning technology to make the robot adjust the grasping mode and path adaptively, which improves the flexibility and success rate of grasping. The experimental results show that the system can maintain high recognition accuracy and success rate in different lighting conditions, object occlusion and complex background. The system has obvious advantages in improving the intelligent level of industrial robots and promoting the development of intelligent manufacturing.

Keywords

Deep learning; Industrial robots; Visual recognition; Grab strategy; Smart manufacturing