CA-I2P: Channel-Adaptive Registration Network with Global Optimal Selection

TL;DR

This paper introduces CA-I2P, a novel registration network that adaptively adjusts feature channels and globally optimizes correspondence selection, significantly improving image-to-point cloud registration accuracy.

Contribution

The paper proposes CAA and GOS modules to address feature mismatch and redundant correspondences, advancing registration methods with global optimization techniques.

Findings

Achieves state-of-the-art registration accuracy on RGB-D Scenes V2.

Outperforms existing methods in handling cross-modal feature differences.

Demonstrates robustness in scene matching with complex structures.

Abstract

Detection-free methods typically follow a coarse-to-fine pipeline, extracting image and point cloud features for patch-level matching and refining dense pixel-to-point correspondences. However, differences in feature channel attention between images and point clouds may lead to degraded matching results, ultimately impairing registration accuracy. Furthermore, similar structures in the scene could lead to redundant correspondences in cross-modal matching. To address these issues, we propose Channel Adaptive Adjustment Module (CAA) and Global Optimal Selection Module (GOS). CAA enhances intra-modal features and suppresses cross-modal sensitivity, while GOS replaces local selection with global optimization. Experiments on RGB-D Scenes V2 and 7-Scenes demonstrate the superiority of our method, achieving state-of-the-art performance in image-to-point cloud registration.