RICAIP Training

An approach for precise detection of uneven surfaces based on Best-Fit, Iterative Closest Point and Angle Criterion methods

The Training/Tutorial will introduce an approach for precise detection of uneven surfaces of objects based on improved point cloud registration methods. The uneven surfaces are deviations between intended (target) and actual (real) geometries, which inevitably occur during the primary molding process: casting.

In this training, we will explore the approach that utilizes cloud imaging from both sanded (desired) and unsanded (real) objects to identify deviations from intended geometries. We will show how together best-fit and Iterative Closest Point methods enhance the accuracy of detection. And how Angle Criterion method implemented to filtered out the edge noises resulting from misalignment in the point cloud. We will cover also the transformation of the point cloud data representations of uneven surfaces between the camera and robot, necessary for the sanding path planning.

Agenda

• Introduction
  • Sanding process
  • Challenge/Problem
  • Main Contribution

• Precise Detection of Uneven Surfaces: Overview and Concept
  • Best-Fit Method for ICP initial transformation matrix estimation
  • Iterative ICP Method for Desired (what) and Real Point Cloud Registration
  • Angle Criterion Method for removing edge noise point-cloud
  • Transformation of the clean uneven surface point from Camera Coordinates System to Robot Base Coordinates System

• Evaluation and Validation of the Concept
• Conclusion

Speaker

Register

To ensure your attendance at the RICAIP Training on 22 May 2024, please fill in the following information by Friday May 3rd.

We will then contact you to confirm your participation and any details of the programme.