SPEED+: Next-Generation Dataset for Spacecraft Pose Estimation across Domain Gap

Park, Tae Ha; Märtens, Marcus; Lecuyer, Gurvan; Izzo, Dario; D'Amico, Simone

doi:10.48550/arxiv.2110.03101

Cited by 4 publications

(15 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The dataset labels include quaternion and 3D depth distance. In 2021, they further proposed Next-Generation Dataset for Spacecraft Pose Estimation across Domain Gap, called SPEED+ [35]. In addition to the 60,000 synthetic images used for training, SPEED+ includes 9,531 real images taken from Tron.…”

Section: A Space Target Datasetsmentioning

confidence: 99%

A Diverse Space Target Dataset With Multidebris and Realistic On-Orbit Environment

Zhang

Deng

2022

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

As the number of space targets increases year by year, space situational awareness plays an important role in satellite security. To keep satellite secure, we use space target monitoring, removal and avoidance to complete space situational awareness. These tasks usually use space target classification algorithms, and the algorithms should be compared fairly with a space target dataset. But at present, real data collection is difficult, and the existing simulation datasets lack some space models. These incomplete space target datasets limit the sustainable development of this field. Therefore, this paper proposes a space target dataset to make up for the shortcomings of the existing space target dataset. (1) More realistic. Our dataset considers the real distribution and operation of space targets and counts their real number and orbital position. We simulate the real space target environment to ensure that the dataset can be migrated to practical applications. ( 2) More diverse. Our dataset includes 11 classes of satellites and 35 classes of space debris. In addition, we use a large number of lighting angles and shooting orbits for each model. (3) More reliable. In order to reduce artificial data bias, we count the attributes of the dataset to analyze the consistent distribution of attributes such as the resolution and contrast of space target images in different classes. Then we made some targeted modifications. In addition, we propose a feature-based quantitative evaluation method for reliability, which calculates the distribution among different classes in the space target dataset.

show abstract

Section: A Space Target Datasetsmentioning

confidence: 99%

A Diverse Space Target Dataset With Multidebris and Realistic On-Orbit Environment

Zhang

Deng

2022

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

show abstract

“…In particular, the Cygnus dataset [4] contains 540 pictures of the Cygnus spacecraft in orbit in conjunction with 20k synthetic images generated with Blender [1]. However, the main limitation of spaceborne images is SPARK [20] SPEED [15] SPEED+ [21] URSO [26] SwissCube [13] Cygnus [4] Prisma12K [22] Prisma25 [7] the lack of accurate pose labels and their limited diversity in terms of pose distribution. To overcome these difficulties, laboratory setups trying to mimic space conditions currently represent the de facto target domain for spacecraft pose estimation algorithms [21].…”

Section: Introductionmentioning

confidence: 99%

“…However, the main limitation of spaceborne images is SPARK [20] SPEED [15] SPEED+ [21] URSO [26] SwissCube [13] Cygnus [4] Prisma12K [22] Prisma25 [7] the lack of accurate pose labels and their limited diversity in terms of pose distribution. To overcome these difficulties, laboratory setups trying to mimic space conditions currently represent the de facto target domain for spacecraft pose estimation algorithms [21]. Moreover, laboratories offer monitoring mockup poses and environmental conditions to ensure a higher quality of the data [30,23].…”

Section: Introductionmentioning

confidence: 99%

“…Therefore, the ongoing second edition of the SPEC will rank the pose estimation algorithms based on their performance on two laboratory datasets, while training images were generated synthetically. That so-called SPEED+ dataset [21] is the first of its kind for vision-only spacecraft pose estimation that combines the information from 60k synthetic images with 10k others acquired from a robotic laboratory setup. However, laboratory pose labels are not publicly available.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

CubeSat-CDT: A Cross-Domain Dataset for 6-DoF Trajectory Estimation of a Symmetric Spacecraft

Musallam

Rathinam

Gaudillière

et al. 2023

Lecture Notes in Computer Science

View full text Add to dashboard Cite

This paper introduces a new cross-domain dataset, CubeSat-CDT, that includes 21 trajectories of a real CubeSat acquired in a laboratory setup, combined with 65 trajectories generated using two rendering engines -i.e. Unity and Blender. The three data sources incorporate the same 1U CubeSat and share the same camera intrinsic parameters. In addition, we conduct experiments to show the characteristics of the dataset using a novel and efficient spacecraft trajectory estimation method, that leverages the information provided from the three data domains. Given a video input of a target spacecraft, the proposed end-to-end approach relies on a Temporal Convolutional Network that enforces the inter-frame coherence of the estimated 6-Degree-of-Freedom spacecraft poses. The pipeline is decomposed into two stages; first, spatial features are extracted from each frame in parallel; second, these features are lifted to the space of camera poses while preserving temporal information. Our results highlight the importance of addressing the domain gap problem to propose reliable solutions for close-range autonomous relative navigation between spacecrafts. Since the nature of the data used during training impacts directly the performance of the final solution, the CubeSat-CDT dataset is provided to advance research into this direction.

show abstract

“…Related studies presented in [ 34 , 35 , 36 ] showed that CNNs can simplify the feature detection and matching process while increasing estimation accuracy. Furthermore, when using CNNs for pose estimation, reducing the domain gap between space imageries and synthetically generated images is another critical topic to be considered [ 45 , 46 , 47 ].…”

Section: Introductionmentioning

confidence: 99%

Monocular Pose Estimation of an Uncooperative Spacecraft Using Convexity Defect Features

Han

Kim

Bang

2022

Sensors

View full text Add to dashboard Cite

Spacecraft relative pose estimation for an uncooperative spacecraft is challenging because the target spacecraft neither provides sensor information to a chaser spacecraft nor contains markers that assist vision-based navigation. Moreover, the chaser does not have prior pose estimates when initiating the pose estimation. This paper proposes a new monocular pose estimation algorithm that addresses these issues in pose initialization situations for a known but uncooperative target spacecraft. The proposed algorithm finds convexity defect features from a target image and uses them as cues for matching feature points on the image to the points on the known target model. Based on this novel method for model matching, it estimates a pose by solving the PnP problem. Pose estimation simulations are carried out in three test scenarios, and each assesses the estimation accuracy and initialization performance by varying relative attitudes and distances. The simulation results show that the algorithm can estimate the poses of spacecraft models when a solar panel length and the number of solar panels are changed. Furthermore, a scenario considering the surface property of the spacecraft emphasizes that robust feature detection is essential for accurate pose estimation. This algorithm can be used for proximity operations with a known but uncooperative target spacecraft. Specifically, one of the main applications is relative navigation for on-orbit servicing.

show abstract

SPEED+: Next-Generation Dataset for Spacecraft Pose Estimation across Domain Gap

Cited by 4 publications

References 30 publications

A Diverse Space Target Dataset With Multidebris and Realistic On-Orbit Environment

A Diverse Space Target Dataset With Multidebris and Realistic On-Orbit Environment

CubeSat-CDT: A Cross-Domain Dataset for 6-DoF Trajectory Estimation of a Symmetric Spacecraft

Monocular Pose Estimation of an Uncooperative Spacecraft Using Convexity Defect Features

Contact Info

Product

Resources

About