HYPER: A New Approach for the Recognition and Positioning of Two-Dimensional Objects.

Area of Science:

• Computer Vision
• Robotics
• Industrial Automation

Background:

• Object identification and localization are crucial for automated tasks.
• Partial occlusions present a significant challenge in real-world industrial environments.
• Existing methods often struggle with variations in lighting, shadows, and object overlap.

Purpose of the Study:

• To develop a robust method for identifying and locating objects on a flat surface.
• To address the challenges posed by partial occlusions in object recognition.
• To enable automated picking and repositioning of industrial parts, even when overlapping.

Main Methods:

• Utilizing a local and compact description of object boundaries.
• Implementing a novel fast recognition method called HYPER (HYpotheses Predicted and Evaluated Recursively).
• Integrating the method into a vision system coupled with an industrial robot arm.

Main Results:

• The HYPER method demonstrates strong robustness to partial occlusions caused by various factors.
• The system successfully performs automatic picking and repositioning of partially overlapping industrial parts.
• The approach provides accurate object identification and localization in challenging conditions.

Conclusions:

• The developed method offers a robust solution for object recognition in the presence of occlusions.
• The integration with robotic systems facilitates automated industrial part handling.
• This approach enhances the capabilities of vision systems in industrial automation.