NLiPsCalib: An Efficient Calibration Framework for High-Fidelity 3D Reconstruction of Curved Visuotactile Sensors

TL;DR

NLiPsCalib is presented, a physics-consistent and efficient calibration framework for curved visuotactile sensors that integrates controllable near-field light sources and leverages Near-Light Photometric Stereo to estimate contact geometry, simplifying calibration to just a few simple contacts with everyday objects.

Abstract

Recent advances in visuotactile sensors increasingly employ biomimetic curved surfaces to enhance sensorimotor capabilities. Although such curved visuotactile sensors enable more conformal object contact, their perceptual quality is often degraded by non-uniform illumination, which reduces reconstruction accuracy and typically necessitates calibration. Existing calibration methods commonly rely on customized indenters and specialized devices to collect large-scale photometric data, but these processes are expensive and labor-intensive. To overcome these calibration challenges, we present NLiPsCalib, a physics-consistent and efficient calibration framework for curved visuotactile sensors. NLiPsCalib integrates controllable near-field light sources and leverages Near-Light Photometric Stereo (NLiPs) to estimate contact geometry, simplifying calibration to just a few simple contacts with everyday objects. We further introduce NLiPsTac, a controllable-light-source tactile sensor developed to validate our framework. Experimental results demonstrate that our approach enables high-fidelity 3D reconstruction across diverse curved form factors with a simple calibration procedure. We emphasize that our approach lowers the barrier to developing customized visuotactile sensors of diverse geometries, thereby making visuotactile sensing more accessible to the broader community.

Figures (8)

• Figure 1: NLiPsCalib, a new calibration pipeline for curved visuotactile sensors that requires no specialized devices. It leverages casual presses with everyday objects, followed by near-light photometric stereo to obtain accurate geometries for building the calibration dataset. This dataset further enables training a neural network for real-time normal inference.
• Figure 2: System pipeline. Using the proposed NLiPsTac tactile sensor, the framework collects a calibration dataset with NLiPs, enabling the training of NLiPsNet, a network designed for real-time 3D shape inference under trichromatic illumination.
• Figure 3: Optical path of near-light photometric stereo calibration, showing illumination from embedded point light sources to the sensor surface, which is then observed by the camera. Controllable LED point light sources are arranged in a ring ($S_0\!-\!S_k$) with known coordinates, and illumination from a selected source (e.g., $S_1$) defines the shaded region.
• Figure 4: Fabrication of the NLiPsTac sensor, including (a) casting a clear elastomer for light transmission and contact support, (b) assembling the 3D-printed base with the LED board, and (c) casting a coating layer to form the reflective surface.
• Figure 5: Normal estimation results for sphere and cube indenters. Columns show Indenter, Pressed, Normals, GT (zoomed), Normals (zoomed), and Error Map. The shared colorbar maps blue $\rightarrow$ red to errors from $0^\circ$ to $25^\circ$.