Curve Set Feature-Based Robust and Fast Pose Estimation Algorithm.

Area of Science:

• Robotics and Automation
• Computer Vision
• Geometric Deep Learning

Background:

• Robotic bin picking is crucial for automated assembly lines, requiring efficient object recognition and localization.
• Current pose estimation algorithms often lack the speed and robustness needed for high-productivity industrial tasks.
• Accurate localization of randomly piled objects from point cloud data remains a significant challenge.

Purpose of the Study:

• To develop a fast and robust pose estimation algorithm for robotic bin picking using point cloud data.
• To introduce novel descriptors and matching techniques for improved performance in cluttered environments.
• To enhance the efficiency and accuracy of object detection and localization in automated manufacturing.

Main Methods:

• Proposed a novel descriptor, Curve Set Feature (CSF), to characterize surface fluctuations around a point for pose evaluation.
• Introduced Rotation Match Feature (RMF) for efficient matching of CSF, combining Point Pair Feature (PPF) concepts with nearest neighbor search.
• Developed a voxel-based pose verification method, significantly outperforming traditional kd-tree approaches in speed.

Main Results:

• The proposed algorithm demonstrates robustness to noise and the ability to detect metal parts.
• Achieved significantly higher accuracy and speed compared to existing methods like PPF and Oriented, Unique and Repeatable (OUR)-Clustered Viewpoint Feature Histogram (CVFH).
• The voxel-based verification is over 30-times faster than kd-tree-based methods, and overall performance is more than 10-times faster than PPF and OUR-CVFH.

Conclusions:

• The developed pose estimation algorithm offers a substantial improvement for robotic bin picking tasks.
• The novel CSF and RMF descriptors, coupled with voxel-based verification, provide a fast, robust, and accurate solution for industrial automation.
• This advancement contributes to increased productivity and efficiency in automated assembly lines.