Markers Identification for Relative Pose Estimation of an Uncooperative Target

TL;DR

The paper tackles autonomous relative pose estimation for de-orbiting uncooperative space objects, focusing on ENVISAT. It integrates CNN-based corner/marker detection (via L-CNN) on chaser imagery with an Unscented Kalman Filter to estimate a $12$-component state vector comprising relative position, velocity, MRPs ${\mathbf{p}}$, and angular rates within LVLH dynamics, using MATLAB simulations and noise/blur pre-processing. A marker-based measurement model leverages ENVISAT corner geometry and accounts for variable marker visibility, enabling robust state updates through the UKF. The results indicate promising corner-detection robustness and estimation accuracy when multiple markers are visible, signaling potential for scalable, autonomous space-debris removal aligned with IADC recommendations.

Abstract

This paper introduces a novel method using chaser spacecraft image processing and Convolutional Neural Networks (CNNs) to detect structural markers on the European Space Agency's (ESA) Environmental Satellite (ENVISAT) for safe de-orbiting. Advanced image pre-processing techniques, including noise addition and blurring, are employed to improve marker detection accuracy and robustness. Initial results show promising potential for autonomous space debris removal, supporting proactive strategies for space sustainability. The effectiveness of our approach suggests that our estimation method could significantly enhance the safety and efficiency of debris removal operations by implementing more robust and autonomous systems in actual space missions.

Figures (9)

• Figure 1: Visual representation of the elements of translational dynamics
• Figure 2: ENVISAT satellite with its geometrical dimensions
• Figure 3: Simulated images of ENVISAT satellite
• Figure 4: Overall CNN architectural diagram
• Figure 5: Preliminary results of the CNN-based corner detection algorithm