Target Recognition of Industrial Robots Using Machine Vision in 5G Environment

Area of Science:

• Robotics
• Computer Vision
• Artificial Intelligence

Background:

• Industrial robots require precise object detection for efficient operation.
• Current systems face challenges with positioning errors, slow recognition, and low accuracy, especially in 5G environments.
• Deep learning (DL) offers potential solutions for enhancing visual recognition capabilities.

Purpose of the Study:

• To address limitations in industrial robot detection, including large positioning errors and low recognition accuracy.
• To optimize industrial robot visual recognition systems using improved deep learning algorithms.
• To evaluate the effectiveness of enhanced Fast-RCNN and VGG-16 models for target detection and classification.

Main Methods:

• Implemented a convolutional neural network (CNN) model for image convolution, pooling, and target classification.
• Utilized an improved Fast-RCNN model for detecting bottled objects.
• Employed an improved VGG-16 classification network with a Hyper-Column scheme for small objects in complex environments.
• Compared simulation results with other advanced CNN algorithms.

Main Results:

• The improved Fast-RCNN and VGG-16 models achieved a recognition accuracy rate of 82.34%.
• Both models demonstrated superior performance in positioning and recognizing targets compared to other advanced CNN algorithms.
• The improved VGG-16 network with the Hyper-Column scheme showed significant accuracy and effectiveness.

Conclusions:

• The improved VGG-16 classification network based on the Hyper-Column scheme provides accurate and effective target recognition and positioning for industrial robots.
• This approach offers a valuable experimental reference for the application and development of industrial robots in 5G environments.
• Enhanced deep learning models can overcome existing challenges in industrial robot visual perception.