Intelligent Perception System of Robot Visual Servo for Complex Industrial Environment

Area of Science:

• Computer Vision
• Robotics
• Machine Learning

Background:

• Industrial robot control relies heavily on visual perception for complex environment navigation.
• Recognizing small or distant targets in cluttered industrial settings poses significant challenges for current systems.
• Accurate target recognition is a critical prerequisite for effective visual servoing.

Purpose of the Study:

• To develop an advanced image processing algorithm for robust target recognition in industrial environments.
• To enhance the capabilities of visual servo systems by improving automatic target identification and classification.
• To address limitations in detecting targets under complex visual backgrounds, varying lighting, and different perspectives.

Main Methods:

• Proposed a novel You Only Look Once Version 2 Region of Interest (YOLO-v2-ROI) neural network algorithm.
• Combined the rapid detection capabilities of YOLO with the Region of Interest (ROI) pooling structure for precise identification.
• Machine learning techniques were employed for training and validating the algorithm.

Main Results:

• The YOLO-v2-ROI algorithm demonstrated real-time image detection speeds.
• Achieved strong adaptability and recognition accuracy in complex backgrounds, including varied lighting and perspectives.
• Successfully identified and located visual targets, improving the overall environmental adaptability of visual servo systems.

Conclusions:

• The proposed YOLO-v2-ROI algorithm effectively overcomes challenges in industrial visual servo systems.
• The method enhances robot vision system efficiency by enabling automatic target recognition and reducing computational load.
• This approach significantly improves the environmental adaptability and performance of robots operating in complex industrial settings.