LAR Detection: Analysis of Classical Approaches, Failures and Mitigation Strategies

Kirimamuni Supuni Kaveendya, Lina Amaya Mejia, Miguel Olivares-Mendez, Carol Martinez

Keywords: Launch Adapter ring(LAR), Edge Detection, On-Orbit Servicing(OOS), Illumination Robustness, Space Robotics

TL;DR: Comparative analysis of classical edge-detection methods for LAR detection in robotic on-orbit servicing, including failure-case analysis and mitigation for efficient, illumination-robust performance.

Abstract:

The Launch Adapter Ring (LAR) acts as both a distinctive visual marker and a structural interface for grasping. Reliable perception of the LAR is critical for successful on-orbit servicing (OOS) missions, where a servicer spacecraft must accurately localize and approach a target vehicle under visually challenging conditions imposed by the harsh space environment. This paper presents a performance analysis of classical edge-based detection methods for Launch Adapter Ring (LAR) detection. We evaluate Sobel, Canny, and Laplacian-of-Gaussian (LoG) edge detectors within a unified perception pipeline, accompanied by a failure cases analysis and a review of mitigation strategies.
Experiments on representative datasets demonstrate the overall robustness of edge-based approaches, with the proposed pipeline achieving a high average detection accuracy of 0.94 across inner and outer rings (Canny: 1.00, LoG: 0.85, Sobel: 0.92). However, the study also reveals distinct failure cases under harsh illumination conditions, as well as false detections caused by satellite body clutter and specular reflections. Therefore, we further review existing mitigation strategies from both classical CV and DL perspectives, with particular attention to computational efficiency. The results highlight the strengths and limitations of classical edge detection for LAR perception in OOS scenarios and demonstrate that the YOLOv8n + phase congruency hybrid provides the most computationally efficient and illumination-robust solution.

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Paper Acceptance: No

Submission Number: 23