Rig-Aware 3D Reconstruction of Vehicle Undercarriages using Gaussian Splatting

TL;DR

This paper tackles the challenge of remote, reliable inspection of vehicle undercarriages by capturing multi-view wide-angle video with a three-camera rig and reconstructing photorealistic, interactive 3D models. It introduces a rig-aware Structure-from-Motion pipeline that integrates precise calibration, sub-frame video synchronization, constrained learned feature matching (DISK+LightGlue), and rig priors in bundle adjustment to overcome low parallax and distortion. The sparse reconstruction seeds a fast 3D Gaussian splatting stage that yields real-time renderable radiance representations, achieving high fidelity (PSNR $>30$ dB, SSIM $>0.9$, LPIPS $<0.2$) and practical speeds ($>130$ FPS). The approach enables safer inspections, increased buyer confidence, and scalable deployment in marketplaces and maintenance workflows, with potential extensions to automated damage detection and other industrial inspection tasks.

Abstract

Inspecting the undercarriage of used vehicles is a labor-intensive task that requires inspectors to crouch or crawl underneath each vehicle to thoroughly examine it. Additionally, online buyers rarely see undercarriage photos. We present an end-to-end pipeline that utilizes a three-camera rig to capture videos of the undercarriage as the vehicle drives over it, and produces an interactive 3D model of the undercarriage. The 3D model enables inspectors and customers to rotate, zoom, and slice through the undercarriage, allowing them to detect rust, leaks, or impact damage in seconds, thereby improving both workplace safety and buyer confidence. Our primary contribution is a rig-aware Structure-from-Motion (SfM) pipeline specifically designed to overcome the challenges of wide-angle lens distortion and low-parallax scenes. Our method overcomes the challenges of wide-angle lens distortion and low-parallax scenes by integrating precise camera calibration, synchronized video streams, and strong geometric priors from the camera rig. We use a constrained matching strategy with learned components, the DISK feature extractor, and the attention-based LightGlue matcher to generate high-quality sparse point clouds that are often unattainable with standard SfM pipelines. These point clouds seed the Gaussian splatting process to generate photorealistic undercarriage models that render in real-time. Our experiments and ablation studies demonstrate that our design choices are essential to achieve state-of-the-art quality.

Figures (9)

• Figure 1: 3D reconstruction pipeline overview. (1) We perform a one-time camera calibration using a ChArUco board to model and correct for severe wide-angle lens distortion. (2) For each vehicle, we synchronize the raw videos from the three cameras on our rig to ensure spatial-temporal alignment. (3) From the synchronized videos, we uniformly sample the sharpest frames triplets, undistort the frames, extract DISK features, apply a constrained feature matching strategy to find matches across different frames via LightGlue, and generate the 3D sparse point cloud of the undercarriage via bundle adjustment. (4) Finally, the point cloud is used to initialize the 3D Gaussians, which are optimized to produce the interactive 3D undercarriage model.
• Figure 2: Sample of ChArUco board opencv_charuco_tutorial.
• Figure 3: Video synchronization of two videos using phase correlation and L1 loss minimization.
• Figure 4: Qualitative comparison of original calibration image (left) and undistorted image using Full OpenCV model (right).
• Figure 5: Qualitative comparison of original undercarriage image (left) and undistorted image using Full OpenCV model (right).