Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

PredRecon: A Prediction-boosted Planning Framework for Fast and High-quality Autonomous Aerial Reconstruction

Chen Feng, Haojia Li, Fei Gao, Boyu Zhou, Shaojie Shen

TL;DR

PredRecon addresses inefficiencies in autonomous aerial 3D reconstruction by introducing a Surface Prediction Module that infers complete surfaces from partial observations, enabling a prediction-guided global coverage. A hierarchical planner then produces a global path and a locally optimized, MVS-aware trajectory to maximize reconstruction quality with a single flight. The approach combines end-to-end surface prediction, online volumetric mapping, and ATSP-based global planning with local Dijkstra-based optimization, validated in realistic simulations against state-of-the-art methods and demonstrated with real-time onboard planning. Open-source code is released to facilitate adoption and replication in practical deployments.

Abstract

Autonomous UAV path planning for 3D reconstruction has been actively studied in various applications for high-quality 3D models. However, most existing works have adopted explore-then-exploit, prior-based or exploration-based strategies, demonstrating inefficiency with repeated flight and low autonomy. In this paper, we propose PredRecon, a prediction-boosted planning framework that can autonomously generate paths for high 3D reconstruction quality. We obtain inspiration from humans can roughly infer the complete construction structure from partial observation. Hence, we devise a surface prediction module (SPM) to predict the coarse complete surfaces of the target from the current partial reconstruction. Then, the uncovered surfaces are produced by online volumetric mapping waiting for observation by UAV. Lastly, a hierarchical planner plans motions for 3D reconstruction, which sequentially finds efficient global coverage paths, plans local paths for maximizing the performance of Multi-View Stereo (MVS), and generates smooth trajectories for image-pose pairs acquisition. We conduct benchmarks in the realistic simulator, which validates the performance of PredRecon compared with the classical and state-of-the-art methods. The open-source code is released at https://github.com/HKUST-Aerial-Robotics/PredRecon.

PredRecon: A Prediction-boosted Planning Framework for Fast and High-quality Autonomous Aerial Reconstruction

TL;DR

PredRecon addresses inefficiencies in autonomous aerial 3D reconstruction by introducing a Surface Prediction Module that infers complete surfaces from partial observations, enabling a prediction-guided global coverage. A hierarchical planner then produces a global path and a locally optimized, MVS-aware trajectory to maximize reconstruction quality with a single flight. The approach combines end-to-end surface prediction, online volumetric mapping, and ATSP-based global planning with local Dijkstra-based optimization, validated in realistic simulations against state-of-the-art methods and demonstrated with real-time onboard planning. Open-source code is released to facilitate adoption and replication in practical deployments.

Abstract

Autonomous UAV path planning for 3D reconstruction has been actively studied in various applications for high-quality 3D models. However, most existing works have adopted explore-then-exploit, prior-based or exploration-based strategies, demonstrating inefficiency with repeated flight and low autonomy. In this paper, we propose PredRecon, a prediction-boosted planning framework that can autonomously generate paths for high 3D reconstruction quality. We obtain inspiration from humans can roughly infer the complete construction structure from partial observation. Hence, we devise a surface prediction module (SPM) to predict the coarse complete surfaces of the target from the current partial reconstruction. Then, the uncovered surfaces are produced by online volumetric mapping waiting for observation by UAV. Lastly, a hierarchical planner plans motions for 3D reconstruction, which sequentially finds efficient global coverage paths, plans local paths for maximizing the performance of Multi-View Stereo (MVS), and generates smooth trajectories for image-pose pairs acquisition. We conduct benchmarks in the realistic simulator, which validates the performance of PredRecon compared with the classical and state-of-the-art methods. The open-source code is released at https://github.com/HKUST-Aerial-Robotics/PredRecon.
Paper Structure (17 sections, 16 equations, 6 figures, 3 tables)

This paper contains 17 sections, 16 equations, 6 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Surface Prediction and Completion
  4. Path Planning for Aerial Reconstruction
  5. System Overview
  6. Surface Prediction Module
  7. Data Pre-process
  8. Prediction Network Structure
  9. Volumetric Mapping with Prediction
  10. Hierarchical Planner
  11. Global Coverage Path Planning
  12. Quality-driven Local Path Planning
  13. Experiments
  14. Implementation Details
  15. Benchmark Comparisons
  16. ...and 2 more sections

Figures (6)

  • Figure 1: (a) Illustration of the proposed framework results during executing trajectory for 3D reconstruction, (b) 3D reconstruction result of the above target produced by the proposed framework.
  • Figure 2: The overview of the proposed prediction-boosted path planning framework for 3D reconstruction.
  • Figure 3: The overall architecture of the proposed SPM (Sect.\ref{['sec:spm']}).
  • Figure 4: Global coverage path Planning: (1) Cluster the uncovered surfaces. (2) Viewpoints generation through dual sampling. (3) The global coverage path is given by the ATSP solver. (Sect.\ref{['subs:global']})
  • Figure 5: Quality-driven local path planning based on the graph search. Through fully considering MVS-related factors, a reconstruction quality-driven local path is produced with its corresponding trajectory. (Sect.\ref{['subs:local']})
  • ...and 1 more figures