How semantic and geometric information mutually reinforce each other in ToF object localization

TL;DR

This paper introduces a two-CNN approach that leverages semantic and geometric information from ToF sensor data to improve 3D object localization accuracy, outperforming conventional methods.

Contribution

The novel integration of calibration and ground-referenced geometric features into CNN-based segmentation enhances localization precision in ToF sensor data.

Findings

Significant improvement in segmentation accuracy.

Enhanced object localization accuracy over baseline methods.

Effective use of ground-referenced geometric features.

Abstract

We propose a novel approach to localize a 3D object from the intensity and depth information images provided by a Time-of-Flight (ToF) sensor. Our method uses two CNNs. The first one uses raw depth and intensity images as input, to segment the floor pixels, from which the extrinsic parameters of the camera are estimated. The second CNN is in charge of segmenting the object-of-interest. As a main innovation, it exploits the calibration estimated from the prediction of the first CNN to represent the geometric depth information in a coordinate system that is attached to the ground, and is thus independent of the camera elevation. In practice, both the height of pixels with respect to the ground, and the orientation of normals to the point cloud are provided as input to the second CNN. Given the segmentation predicted by the second CNN, the object is localized based on point cloud alignment with a reference model. Our experiments demonstrate that our proposed two-step approach improves segmentation and localization accuracy by a significant margin compared to a conventional CNN architecture, ignoring calibration and height maps, but also compared to PointNet++.