Correspondence Matrices are Underrated

TL;DR

This paper argues that using correspondence error as a loss function in point-cloud registration improves convergence speed and accuracy, especially in cases of larger misalignment, by focusing on point match quality.

Contribution

The paper introduces the idea of using correspondence error as a loss function in deep-learning-based point-cloud registration methods, demonstrating its advantages over traditional transformation error-based losses.

Findings

Modified existing methods with correspondence-based loss converge faster.

Correspondence-based loss improves registration accuracy.

Methods perform better at larger misalignments.

Abstract

Point-cloud registration (PCR) is an important task in various applications such as robotic manipulation, augmented and virtual reality, SLAM, etc. PCR is an optimization problem involving minimization over two different types of interdependent variables: transformation parameters and point-to-point correspondences. Recent developments in deep-learning have produced computationally fast approaches for PCR. The loss functions that are optimized in these networks are based on the error in the transformation parameters. We hypothesize that these methods would perform significantly better if they calculated their loss function using correspondence error instead of only using error in transformation parameters. We define correspondence error as a metric based on incorrectly matched point pairs. We provide a fundamental explanation for why this is the case and test our hypothesis by modifying existing methods to use correspondence-based loss instead of transformation-based loss. These experiments show that the modified networks converge faster and register more accurately even at larger misalignment when compared to the original networks.