Pith. sign in

REVIEW 3 major objections 4 minor 51 references

Revisiting-Aware In-Orbit Edge Computing for Earth Observation

T0 review · 3 major / 4 minor · reviewed 2026-08-01 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Earth observation satellites can deliver up to 4.55x more revisiting imagery under the same downlink budget by storing past images onboard and transmitting only the pixels that changed.

desk verdict Stride's core idea is genuinely worth a look, but the headline gains rest on a 24-hour revisit assumption that doesn't hold for the very satellites it cites. read the letter →

arxiv 2607.25813 v1 pith:JAGCUSV6 submitted 2026-07-28 cs.NI

Zehua Sun , Tao Ni , Kaiyan Cui , Weitao Xu , Jingxian Wang This is my paper · ORCID
classification cs.NI
keywords Earthobservationsatellitesrevisitingcyclein-orbitedgecomputingtemporalredundancychangedetectionsatellitedownlinkbottleneckimagecompressionregionsofinterest
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper argues that Earth observation satellites waste their scarce downlink capacity by re-sending imagery they have already captured on previous passes. It proposes Stride, an onboard pipeline that keeps a historical reference image of each revisited area, aligns the new capture to that reference, detects only the pixels that actually changed, and sends only those regions; the ground station reconstructs the full image by adding the differences back to the reference. The paper reports that this raises the fraction of revisiting imagery delivered (the RID score) by up to 4.55x, cuts connectivity latency by 5.02x, and enlarges mapping coverage by 2.56x while keeping reconstruction quality around SSIM 0.99. If those numbers hold in orbit, existing constellations could observe the same areas more often without new satellites or additional ground stations.

What carries the argument

The load-bearing object is the sparse differencing matrix ΔI = I − I_ref, computed after aligning the new image to a stored revisiting reference. Three mechanisms protect its sparsity: a mono- and multi-temporal cloud indicator rejects cloudy captures, a coarse-to-fine reference selector aligns images via SURF keypoints and global-consensus origin voting to correct orbit deviation, and an ensemble-local change detector suppresses pixel-level noise and encodes the surviving Regions of Interest in Compressed Sparse Row format. The machinery's role is to convert a downlink-bandwidth problem into a sparsity problem: if the difference matrix is sparse, the ground station reconstructs a full-quali

What would settle it

Take a real revisit pair of the same area separated by a season change or a large sun-angle change, or with a cloudy stored reference, and run Stride's full pipeline on a flight-like computer. Measure the size ratio of the transmitted sparse matrix and the SSIM after ground reconstruction: if the size ratio stays near 1.0 for a substantial share of revisits, or if enforcing the nominal reference collapses SSIM below the reported 0.99, the central mechanism fails.

Watch

Extended reading notes

Core claim

Stride's central claim is that a satellite's revisiting cycle is itself a compression resource: because the same ground area is imaged repeatedly, the onboard computer can hold a reference image from a previous visit and transmit only the delta. The paper presents this as the first system to exploit orbital revisiting properties for in-orbit compression, and argues that the combination of a mono- and multi-temporal cloud indicator, a coarse-to-fine reference selector, and an ensemble-local change detector makes the delta sparse enough in practice. In the reported experiments, Stride reaches a Revisiting Imagery Delivery score of 0.8776 for SKYSAT-A against 0.2982 for raw delivery, while main

Load-bearing premise

The entire gain rests on the assumption that a newly captured image of the same area, after geometric alignment, differs from the stored reference only at a small, sparse set of pixels—i.e., that temporal redundancy is high enough, and the reference is clear enough, for the difference matrix to be mostly zeros.

Editorial extensions

If this is right

  • The same constellation and ground-station layout can deliver nearly the full revisiting record within a 24-hour window instead of roughly a third of it, shortening the perceived revisit period.
  • Per-image downlink load can drop to about a quarter of the raw image at near-lossless quality, making low-latency monitoring of fast-changing areas feasible with existing downlink capacity.
  • Onboard storage need only exceed roughly one raw image per revisited area, well within current multi-terabyte satellite storage, so the approach does not require a new hardware class.
  • If cloud detection fails or no valid reference exists, the image is transmitted directly, so the system degrades to raw delivery rather than losing data.
  • In the reported comparisons, Stride transmits less than either lossless conventional compression or reference-uplink compression at comparable reconstruction quality, because it exploits temporal redundancy without requiring ground-to-satellite reference uploads.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The same reference-and-delta loop could be viewed as an open-loop video codec for Earth observation; a natural extension is adaptive reference refresh, where the satellite re-broadcasts a fresh reference whenever accumulated change or drift makes the delta dense.
  • Since the gains depend on clear-sky similarity, constellation scheduling could be designed to pair revisits with similar sun-sensor geometry and to avoid storing cloudy references, turning Stride's sensitivity to weather into a planning mechanism.
  • The core principle—spend onboard storage instead of downlink bandwidth by exploiting repeat ground tracks—could generalize beyond imaging, for example to onboard sensor time series or to inter-satellite sharing of references when two satellites see the same area with compatible geometry.
  • A radiometric normalization step (e.g., compensating for sun angle and seasonal reflectance) before differencing would be a direct testable extension; without it, the sparsity assumption is most vulnerable exactly when images are captured under changing illumination.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper proposes Stride, an onboard edge-computing framework for Earth-observation satellites that exploits temporal redundancy in revisiting imagery. For each captured image, Stride performs cloud detection, selects a stored historical reference image via SURF-based alignment, and transmits only the regions that differ from the reference, encoded as a sparse matrix; the ground station reconstructs the full image using its own copy of the reference. The authors report a Flat-Sat implementation on a Jetson TX2 and a constellation simulator using TLE descriptors for Landsat-8, Sentinel-2A, and SKYSAT-A. They claim up to a 4.55× improvement in a Revisiting Imagery Delivery (RID) score, 5.02× lower connectivity latency, 2.56× larger mapping coverage, and state-of-the-art compression relative to JPEG-2000, CCSDS 121.0-B-3, and Earth+, while maintaining reconstruction quality around SSIM≈0.99.

Significance. If the claimed gains hold, the paper makes a useful contribution: it is, to my knowledge, the first system to avoid the uplink dependency of reference-based compression by storing historical revisiting imagery onboard, and the Flat-Sat measurements provide credible evidence that the compute overhead is modest for a power-constrained platform. The use of external baselines (JPEG-2000, CCSDS, Earth+) rather than only self-comparison is a strength, and the RID metric, although defined by the authors, is naturally tied to the downlink bottleneck the paper targets. However, the central claim depends on two load-bearing assumptions that the current manuscript does not sufficiently validate: that the tested orbits actually provide same-area references within the simulated 24-hour window, and that the reported compression ratios are achieved at comparable reconstruction quality to the baselines in the 'lossless' comparison. Both issues are addressable with additional experiments and clarification, but they affect the credibility of the headline multipliers.

major comments (3)
  1. [Section V and Section IV-C2] The simulator time base is inconsistent with the real repeat cycles of the tested satellites. Section V states that the simulator operates during 'one revisiting cycle (i.e., 24 hours)' for Landsat-8, Sentinel-2A, and SKYSAT-A, and Figure 11a computes RID over 24 hours. However, Landsat-8 has a 16-day exact-repeat cycle and Sentinel-2A a 10-day cycle for a single satellite. Without explicit off-nadir/adjacent-orbit scheduling, a same-area reference image would not be available on a daily cadence for these satellites. Since Stride's RoI mechanism requires a stored historical reference (Section IV-D, Eq. (5)), the reported RID/latency/coverage gains for these two satellites are only valid if the simulator supplies near-daily references that the actual orbits cannot provide. Only geometric SURF alignment is applied (Section IV-C2); no radiometric normalization is described. Please rerun the
  2. [Section VI-F, Figure 16a] The 'lossless settings' comparison is not apples-to-apples. Stride's local compensation (Eq. (5)) zeros all pixels below the threshold τ, and the result is sparse-encoded via CSR (Eq. (6)); this is a lossy operation for τ>0. JPEG-2000 and CCSDS are evaluated in lossless mode with mean SSIM=1.00, but the figure reports Stride's size ratio (0.22) under the same 'lossless setting' without reporting its SSIM or MAE. This conflates rate and distortion: the gap between 0.22 and 0.40/0.51 may be partly a quality trade-off. Please report rate-distortion curves (size ratio vs SSIM/MAE) for Stride and the baselines across τ, or clearly state the lossy operating point used in Stride's comparison.
  3. [Section VI-A, Figures 10-11] The headline 4.55× RID improvement is quoted at τ=0.2, the most aggressive threshold tested. At this setting the size ratio is 0.22, but Section VI-A also states that increasing τ lowers reconstruction quality (Figure 10c-d), and no SSIM/MAE value is given for the exact operating point used in Figure 11a. The claim of 'maintaining reconstruction quality' is therefore not substantiated at the headline operating point. In addition, no error bars or confidence intervals are reported for the RID/size-ratio numbers, even though Section VI-F says the quantitative comparison was run 'under part of random regions.' Please report the quality metrics at τ=0.2 and the variance across random subsets.
minor comments (4)
  1. [Section III] The RID score definition would benefit from a precise formula. As written, 'ratio of successfully delivered data to total collected data' is ambiguous: does the numerator count raw-equivalent pixels after decompression, or transmitted bytes? If it counts raw-equivalent pixels, clarify that RID can exceed the raw delivery fraction when compression is used; if it counts transmitted bytes, the metric is trivially improved by any compression scheme and needs a different interpretation.
  2. [Abstract and Section IV-A] The text says Stride transmits 'only the RoIs,' but Figure 10a explicitly includes direct transmission of cloudy images, reference images, and overloaded data. Suggest using a phrase such as 'RoI-centric' or 'RoI-first' to avoid overclaiming.
  3. [Equation (5)] The notation max(I(h,w,0)) is ambiguous. It should be written as max_{h,w} I(h,w,0), i.e., the global maximum over the first channel of the candidate image, to make the threshold definition unambiguous.
  4. [Section V] The default parameter 'epsilon=1e7' is listed in the default parameters paragraph but is not used in any subsequent equation or evaluation description. Either remove it or define its role in the reference selector.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: Stride's RID/latency/coverage gains are benchmarked consequences of measured compression ratios, not fitted predictions; the only self-citation is contextual.

full rationale

The paper's central derivation is that transmitting only RoIs reduces per-image transmission load, and the simulator converts that load reduction into a higher RID score (delivered/collected data within 24h). This is an arithmetic consequence, but not circular: RID is an evaluation metric defined in Section III, not a fitted parameter; the compression size ratios (0.22/0.34/0.53 for tau=0.2/0.15/0.1) are measured on public datasets (SpaceNet 7, DynamicEarthNet), and Stride is compared against Raw, JPEG-2000, CCSDS 121.0-B-3, and Earth+ in Figures 10 and 16. No equation defines an output in terms of the quantity it is said to predict, and no parameter is fitted to the RID target and then renamed as a prediction. The local-compensation threshold tau is tuned, and the headline 4.55x uses tau=0.2 (the reciprocal of the 0.22 size ratio), which is a post-hoc threshold choice rather than a circular step; it is a reporting/fairness concern, not a derivation that reduces to its inputs. The only self-citation is [36] (Spacesched) in the related-work survey, which is not load-bearing for Stride's technical claims. The paper itself notes (Section VII, Fault analysis) that if no valid reference exists, the image is transmitted directly, which limits but does not circularize the framework. The skeptic's concern about Landsat-8/Sentinel-2A repeat cycles exceeding the 24h simulator window is an external-validity threat regarding reference availability and cross-cycle similarity, not circularity of the derivation. Therefore no circular step meeting the quoted-evidence standard is present.

Assumptions & free parameters 6 free parameters · 7 assumptions · 0 invented entities

The ledger contains no new physical entities; the system is built from standard image-processing components. The main unstated costs are the hand-set thresholds tau_mono, tau_multi, tau, and the unspecified cloud-cache size n. The domain assumptions about revisit similarity and downlink bottleneck are load-bearing but explicitly stated.

free parameters (6)
  • tau_mono = 0.7 (varied 0.5/0.7)
    Cloud detection threshold in Eq. (1); controls which pixels are flagged as cloud and therefore whether an image is directly transmitted. Tuned per evaluation.
  • tau_multi = 0.7 (varied 0.3/0.5)
    Similarity threshold for multi-temporal cloud verification; hand-set and varied in the cloud indicator evaluation.
  • tau (local compensation threshold) = 0.15 default; 0.1/0.2 swept; headline uses 0.2
    Threshold in Eq. (5) that separates Regions of Interest from Regions of Noise. Directly sets the size ratio and reconstruction quality; the 'up to 4.55x' gain is obtained at the most aggressive value.
  • cloudy classification threshold = 0.25
    Images with cloud coverage score above 0.25 are transmitted directly. Hand-set; no sensitivity analysis.
  • coarse query thresholds tau_lambda, tau_phi = 0.01
    Geographic coordinate ranges used to select candidate reference images. Hand-set defaults.
  • cloud cache size n = not specified
    Number of previous cloud labels stored before triggering multi-temporal verification; the paper gives no concrete value, making the trigger condition partially unspecified.
assumptions (7)
  • domain assumption Satellites periodically revisit the same ground region with similar viewing geometry.
    Central premise of the paper, stated in Section I and Figure 1. If revisit geometry or timing changes too much, the reference images are not comparable.
  • domain assumption Unchanged land surface dominates revisiting imagery, so pixel differencing yields a sparse matrix.
    The whole RoI-transmission scheme relies on this temporal redundancy. Violated by seasonal change, large disasters, or illumination shifts.
  • domain assumption Downlink bandwidth, not onboard storage or compute, is the primary bottleneck.
    The paper cites the 'only 2% captured data received' statistic from prior work [4],[5] and builds the revisiting-cycle-delay problem on it.
  • domain assumption Clouds are brighter than terrain in the RGB channels.
    The mono-temporal cloud detector uses a simple threshold on RGB values (Eq. 1); it does not use multispectral bands such as NIR/SWIR.
  • standard math SURF keypoint matching gives a homography sufficient to align revisit images.
    The fine alignment step in Section IV-C2 assumes enough repeatable keypoints and a planar scene model.
  • domain assumption Ground stations have access to the reference image needed for reconstruction.
    Reconstruction combines transmitted RoIs with a stored reference; the paper counts reference-image transmission in the shadow regions of Figure 10b but does not analyze how a ground station that missed the reference handles the data.
  • domain assumption SSIM/MAE against the original image is the relevant reconstruction-quality metric.
    All quality claims are based on this image-similarity metric; application-level fidelity (e.g., whether a flood extent is correct) is not measured.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Revisiting-Aware In-Orbit Edge Computing for Earth Observation." pith.science (2026). https://pith.science/paper/JAGCUSV6

@misc{pith2026260725813,
  author       = {Pith},
  title        = {Pith review of: Revisiting-Aware In-Orbit Edge Computing for Earth Observation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JAGCUSV6}},
  note         = {Machine review of arXiv:2607.25813}
}
read the original abstract

Typically, Earth observation satellites follow a rule of revisiting cycle to periodically pass over the same area of the Earth at regular intervals, which is jointly determined by their orbital properties (e.g., eccentricity, inclination) and instrument characteristics (e.g., off-nadir pointing and swath capabilities). However, we have observed delays in perceived revisiting cycles where limited satellite downlink bandwidth allows only partial images to be delivered, pushing back the timeliness of the full set of data, which we term as revisiting cycle delay. In this paper, we present a revisiting-aware in-orbit edge computing framework for Earth observation termed Stride. Stride leverages the unique orbital revisiting properties to afford historical reference revisiting images onboard, and exploits the inherent temporal redundancy in the revisiting imagery to transmit only the Regions of Interest (RoIs). Specifically, Stride comprises a mono- and multi-temporal cloud indicator to alleviate cloud contamination, a coarse-to-fine reference selector for orbit deviation correction, and an ensemble-local change detector to address inter-band complexities and pixel-level perturbations. Experiments on a Flat-Sat testbed and a constellation simulator demonstrate Stride improves the Revisiting Imagery Delivery (RID) score by up to 4.55X, decreases the connectivity latency by 5.02X, and enlarges the mapping coverage by 2.56X, yielding state-of-the-art performance.

Figures

Figures reproduced from arXiv: 2607.25813 by the authors.

Figure 1
Figure 1. (a) The revisiting cycle is a primary property of Earth observation satellites, whereby satellites repeatedly pass over the same regions of the Earth. (b) As a satellite revisits a region, the new image often shares substantial temporal redundancy with prior observations. (c) Stride exploits the inherent temporal redundancy among revisiting imagery to transmit only the RoIs, which significantly decreases connec￾tivi… view at source ↗
Figure 4
Figure 4. Illustration of the orbital elements variation over time, [PITH_FULL_IMAGE:figures/full_fig_p003_4.png] view at source ↗
Figure 5
Figure 5. Quantification of revisiting cycle delay. [PITH_FULL_IMAGE:figures/full_fig_p004_5.png] view at source ↗
Figures from the paper (9 more)
Figure 6
Figure 6. Figure 6: Overview of Stride (§ IV-A), consisting of a cloud indicator, a reference selector, and a change detector. groups by constraining its latitude and longitude coordinates within a predefined range (τλ, τϕ). Among them, adjacent potential reference images are concatenated…
Figure 7
Figure 7. Figure 7: Illustration of the change detector on various data [PITH_FULL_IMAGE:figures/full_fig_p005_7.png]
Figure 9
Figure 9. Figure 9: Flat-Sat deployment. TABLE II: Statistics of application datasets. Dataset #Image/ Frame #AOIs Mean GSD (m) Temporal Resolution Time Span Source SpaceNet 7 [2] 1,423 101 4 Monthly 2 years PlanetScope LSCIDMR [29] 104,390 11 - Daily 1 year Himawari-8 DynamicEarthNet [3]…
Figure 10
Figure 10. Figure 10: Overall performance of Stride in terms of transmission load and reconstruction quality. 0 5 10 15 20 Time (hours) 0.0 0.2 0.4 0.6 0.8 1.0 RID score / / / Raw/Stride (SKYSAT-A) Raw/Stride (Sentinel-2A) Raw/Stride (Landsat-8) (a) Revisiting imagery delivery. 0.0 0.2 0.4…
Figure 12
Figure 12. Figure 12: Results of the cloud indicator. and Mean Absolute Error (MAE) to assess the quality of the reconstructed images. (3) Revisiting capability. We utilize the RID score to describe the revisiting capability of satellites. For fair comparison across different satellites, R…
Figure 14
Figure 14. Figure 14: Results of the change detector. 0 10 20 30 40 50 60 70 1400 1500 1600 1700 1800 RAM usage (MB) Idle: 1458.00 MB Stride: 1628.74 MB RAM usage (a) RAM usage. 0 10 20 30 40 50 60 70 0 20 40 60 80 100 CPU usage (%) Idle: 2.99% Stride: 25.67% CPU 1 CPU 2 CPU 3 CPU 4 (b) CP…
Figure 15
Figure 15. Figure 15: System overhead of Stride in terms of RAM usage, CPU load, temperature, and energy consumption. large-size and high-resolution criteria, so these limitations are generally not a concern. D. Evaluation of Change Detector We evaluate Stride’s change detector on the Dyna…
Figure 16
Figure 16. Figure 16: Quantitative comparison with benchmarks (lossless: (a)-(b), lossy: (c)-(d)). [PITH_FULL_IMAGE:figures/full_fig_p009_16.png]
Figure 18
Figure 18. Figure 18: Use case 2: wide-coverage mapping. each reference interval: peaks correspond to reference-image transmission, while valleys occur shortly afterward when im￾age changes are minimal; as time passes, both the changes and the transmission load increase with the varying sp…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

51 extracted references · 1 linked inside Pith

  1. [1]

    Number of active satellites from 1957 to 2022,

    Statista, “Number of active satellites from 1957 to 2022,” https://www. statista.com/statistics/897719/number-of-active-satellites-by-year/, 2023

    1957

  2. [2]

    The spacenet multi-temporal urban development challenge,

    A. Van Etten and D. Hogan, “The spacenet multi-temporal urban development challenge,” arXiv preprint arXiv:2102.11958, 2021

    arXiv 2021

  3. [3]

    Dynam- icearthnet: Daily multi-spectral satellite dataset for semantic change segmentation,

    A. Toker, L. Kondmann, M. Weber, M. Eisenberger, A. Camero, J. Hu, A. P. Hoderlein, C ¸ . S ¸enaras, T. Davis, D. Cremers et al., “Dynam- icearthnet: Daily multi-spectral satellite dataset for semantic change segmentation,” in Proceedings of CVPR, 2022

    2022

  4. [4]

    Orbital edge computing: Nanosatellite con- stellations as a new class of computer system,

    B. Denby and B. Lucia, “Orbital edge computing: Nanosatellite con- stellations as a new class of computer system,” in Proceedings of ACM ASPLOS, 2020

    2020

  5. [5]

    Ko- dan: Addressing the computational bottleneck in space,

    B. Denby, K. Chintalapudi, R. Chandra, B. Lucia, and S. Noghabi, “Ko- dan: Addressing the computational bottleneck in space,” in Proceedings of ACM ASPLOS, 2023

    2023

  6. [6]

    Kepler’s laws of planetary motion: 1609–1666,

    J. L. Russell, “Kepler’s laws of planetary motion: 1609–1666,” The British journal for the history of science, 1964

    1964

  7. [7]

    Repeat ground track orbit design with desired revisit time and optimal tilt,

    M. J. Nadoushan and N. Assadian, “Repeat ground track orbit design with desired revisit time and optimal tilt,” Aerospace Science and technology, 2015

    2015

  8. [8]

    An overview of modis land data processing and product status,

    C. O. Justice, J. R. Townshend, E. F. Vermote, E. Masuoka, R. E. Wolfe, N. Saleous, D. P. Roy, and J. T. Morisette, “An overview of modis land data processing and product status,” RSE, 2002

    2002

Show all 51 references
  1. [9]

    Known knowns and unknowns: Near-realtime earth observation via query bifurcation in serval,

    B. Tao, O. Chabra, I. Janveja, I. Gupta, and D. Vasisht, “Known knowns and unknowns: Near-realtime earth observation via query bifurcation in serval,” in Proceedings of USENIX NSDI, 2024

    2024

  2. [10]

    Jpeg2000: Standard for interactive imaging,

    D. S. Taubman and M. W. Marcellin, “Jpeg2000: Standard for interactive imaging,” Proceedings of the IEEE, 2002

    2002

  3. [11]

    Lossless Data Com- pression,

    Consultative Committee for Space Data Systems, “Lossless Data Com- pression,” https://ccsds.org/Pubs/121x0b3.pdf, 2020

    2020

  4. [12]

    Real- time compressed sensing for joint hyperspectral image transmission and restoration for cubesat,

    C.-C. Hsu, C.-Y . Jian, E.-S. Tu, C.-M. Lee, and G.-L. Chen, “Real- time compressed sensing for joint hyperspectral image transmission and restoration for cubesat,” IEEE TGRS, 2024

    2024

  5. [13]

    Deepspace: Super resolution powered efficient and reliable satellite image data acquistion,

    C. Sun, Y . Zhang, B. Tao, D. Vasisht, and M. Marina, “Deepspace: Super resolution powered efficient and reliable satellite image data acquistion,” in Proceedings of ACM SIGCOMM, 2025

    2025

  6. [14]

    Earth+: On-board satellite imagery compression leveraging historical earth observations,

    K. Du, Y . Cheng, P. Olsen, S. Noghabi, and J. Jiang, “Earth+: On-board satellite imagery compression leveraging historical earth observations,” in Proceedings of ACM ASPLOS, 2025

    2025

  7. [15]

    Cloud and cloud shadow detection for optical satellite imagery: Features, algorithms, validation, and prospects,

    Z. Li, H. Shen, Q. Weng, Y . Zhang, P. Dou, and L. Zhang, “Cloud and cloud shadow detection for optical satellite imagery: Features, algorithms, validation, and prospects,” ISPRS Journal of P&RS, 2022

    2022

  8. [16]

    Improvement and expansion of the fmask algorithm: Cloud, cloud shadow, and snow detection for landsats 4–7, 8, and sentinel 2 images,

    Z. Zhu, S. Wang, and C. E. Woodcock, “Improvement and expansion of the fmask algorithm: Cloud, cloud shadow, and snow detection for landsats 4–7, 8, and sentinel 2 images,” RSE, 2015

    2015

  9. [17]

    Cloud masking for ocean color data processing in the coastal regions,

    M. Wang and W. Shi, “Cloud masking for ocean color data processing in the coastal regions,” IEEE TGRS, 2006

    2006

  10. [18]

    A cloud detection algorithm for satellite imagery based on deep learning,

    J. H. Jeppesen, R. H. Jacobsen, F. Inceoglu, and T. S. Toftegaard, “A cloud detection algorithm for satellite imagery based on deep learning,” RSE, 2019

    2019

  11. [19]

    Cloud detection in remote sensing images based on multiscale features-convolutional neural network,

    Z. Shao, Y . Pan, C. Diao, and J. Cai, “Cloud detection in remote sensing images based on multiscale features-convolutional neural network,” IEEE TGRS, 2019

    2019

  12. [20]

    Orbital decay of satellite galaxies in spherical sys- tems,

    M. D. Weinberg, “Orbital decay of satellite galaxies in spherical sys- tems,” Astrophysical Journal, 1986

    1986

  13. [21]

    Station keeping and momentum management of low-thrust satellites using mpc,

    A. Weiss, U. V . Kalabi ´c, and S. Di Cairano, “Station keeping and momentum management of low-thrust satellites using mpc,” Aerospace Science and Technology, 2018

    2018

  14. [22]

    Change detection methods for remote sensing in the last decade: A comprehensive review,

    G. Cheng, Y . Huang, X. Li, S. Lyu, Z. Xu, H. Zhao, Q. Zhao, and S. Xiang, “Change detection methods for remote sensing in the last decade: A comprehensive review,” Remote Sensing, 2024

    2024

  15. [23]

    Transmitting, fast and slow: Scheduling satellite traffic through space and time,

    B. Tao, M. Masood, I. Gupta, and D. Vasisht, “Transmitting, fast and slow: Scheduling satellite traffic through space and time,” in Proceedings of ACM MobiCom, 2023

    2023

  16. [24]

    Responsive maneuver planning for sun-synchronous repeating ground track orbits,

    S. Treblow and C. N. McGrath, “Responsive maneuver planning for sun-synchronous repeating ground track orbits,” Journal of Spacecraft and Rockets, 2024

    2024

  17. [25]

    Surf: Speeded up robust features,

    H. Bay, T. Tuytelaars, and L. Van Gool, “Surf: Speeded up robust features,” in Proceedings of ECCV, 2006

    2006

  18. [26]

    The role of the hessian matrix in fitting models to measurements,

    W. C. Thacker, “The role of the hessian matrix in fitting models to measurements,” Journal of Geophysical Research: Oceans, 1989

    1989

  19. [27]

    The haar wavelet transform: its status and achievements,

    R. S. Stankovi ´c and B. J. Falkowski, “The haar wavelet transform: its status and achievements,” Computers & Electrical Engineering, 2003

    2003

  20. [28]

    Efficient sparse matrix-vector multipli- cation on gpus using the csr storage format,

    J. L. Greathouse and M. Daga, “Efficient sparse matrix-vector multipli- cation on gpus using the csr storage format,” in Proceedings of ACM SC, 2014

    2014

  21. [29]

    Lscidmr: Large- scale satellite cloud image database for meteorological research,

    C. Bai, M. Zhang, J. Zhang, J. Zheng, and S. Chen, “Lscidmr: Large- scale satellite cloud image database for meteorological research,” IEEE Transactions on Cybernetics, 2021

    2021

  22. [30]

    Mtgl40-5: A multi- temporal dataset for remote sensing image geo-localization,

    J. Ma, S. Pei, Y . Yang, X. Tang, and X. Zhang, “Mtgl40-5: A multi- temporal dataset for remote sensing image geo-localization,” Remote Sensing, 2023

    2023

  23. [31]

    Satsot: A benchmark dataset for satellite video single object tracking,

    M. Zhao, S. Li, S. Xuan, L. Kou, S. Gong, and Z. Zhou, “Satsot: A benchmark dataset for satellite video single object tracking,” IEEE TGRS, 2022

    2022

  24. [32]

    Advancing earth observation: a survey on ai-powered image processing in satellites,

    A. Duggan, B. Andrade, and H. Afli, “Advancing earth observation: a survey on ai-powered image processing in satellites,” European Journal of Remote Sensing, 2025

    2025

  25. [33]

    Satguard: Concealing endless and bursty packet losses in leo satellite networks for delay-sensitive web applications,

    J. Li, H. Li, Z. Lai, Q. Wu, Y . Liu, Q. Zhang, Y . Li, and J. Liu, “Satguard: Concealing endless and bursty packet losses in leo satellite networks for delay-sensitive web applications,” in Proceedings of ACM WWW, 2024

    2024

  26. [34]

    Deciphering the enigma of satellite computing with cots devices: Measurement and analysis,

    R. Xing, M. Xu, A. Zhou, Q. Li, Y . Zhang, F. Qian, and S. Wang, “Deciphering the enigma of satellite computing with cots devices: Measurement and analysis,” in Proceedings of ACM MobiCom, 2024

    2024

  27. [35]

    In-orbit processing or not? sunlight-aware task scheduling for energy-efficient space edge computing networks,

    W. Liu, Z. Lai, Q. Wu, H. Li, Q. Zhang, Z. Li, Y . Li, and J. Liu, “In-orbit processing or not? sunlight-aware task scheduling for energy-efficient space edge computing networks,” in Proceedings of IEEE INFOCOM, 2024

    2024

  28. [36]

    Spacesched: A constellation- wide scheduling system for resolving ground track congestion in remote sensing,

    Z. Sun, T. Ni, P. Hu, T. Gu, and W. Xu, “Spacesched: A constellation- wide scheduling system for resolving ground track congestion in remote sensing,” in Proceedings of ACM MobiCom, 2025

    2025

  29. [37]

    Pmsat: Optimizing passive metasurface for low earth orbit satellite communication,

    H. Pan, L. Qiu, B. Ouyang, S. Zheng, Y . Zhang, Y .-C. Chen, and G. Xue, “Pmsat: Optimizing passive metasurface for low earth orbit satellite communication,” in Proceedings of ACM MobiCom, 2023

    2023

  30. [38]

    Spectrumize: Spectrum-efficient satellite networks for the internet of things,

    V . Singh, T. Chakraborty, S. Jog, O. Chabra, D. Vasisht, and R. Chandra, “Spectrumize: Spectrum-efficient satellite networks for the internet of things,” in Proceedings of USENIX NSDI, 2024

    2024

  31. [39]

    A community-driven approach to democratize access to satellite ground stations,

    V . Singh, A. Prabhakara, D. Zhang, O. Ya ˘gan, and S. Kumar, “A community-driven approach to democratize access to satellite ground stations,” in Proceedings of ACM MobiCom, 2021

    2021

  32. [40]

    Cosmac: Constellation-aware medium access and scheduling for iot satellites,

    J. Shenoy, O. Chabra, T. Chakraborty, S. Jog, D. Vasisht, and R. Chan- dra, “Cosmac: Constellation-aware medium access and scheduling for iot satellites,” in Proceedings of ACM MobiCom, 2024

    2024

  33. [41]

    Satcp: Link-layer informed tcp adaptation for highly dynamic leo satellite networks,

    X. Cao and X. Zhang, “Satcp: Link-layer informed tcp adaptation for highly dynamic leo satellite networks,” in Proceedings of IEEE INFOCOM, 2023

    2023

  34. [42]

    Towards energy-efficient routing in satellite networks,

    Y . Yang, M. Xu, D. Wang, and Y . Wang, “Towards energy-efficient routing in satellite networks,” IEEE JSAC, 2016

    2016

  35. [43]

    Stable hierarchical routing for operational leo networks,

    Y . Li, L. Liu, H. Li, W. Liu, Y . Chen, W. Zhao, J. Wu, Q. Wu, J. Liu, and Z. Lai, “Stable hierarchical routing for operational leo networks,” in Proceedings of ACM MobiCom, 2024

    2024

  36. [44]

    The dark side of scale: Insecurity of direct-to-cell satellite mega-constellations,

    W. Liu, Y . Li, H. Li, Y . Chen, Y . Wang, J. Lan, J. Wu, Q. Wu, J. Liu, and Z. Lai, “The dark side of scale: Insecurity of direct-to-cell satellite mega-constellations,” in Proceedings of IEEE S&P, 2024

    2024

  37. [45]

    Sate: Low- latency traffic engineering for satellite networks,

    H. Wu, Y . Han, M. Rajpal, Q. Zhang, and J. Wang, “Sate: Low- latency traffic engineering for satellite networks,” in Proceedings of ACM SIGCOMM, 2025

    2025

  38. [46]

    Democratizing{Direct-to-Cell}low earth orbit satellite networks,

    L. Liu, Y . Li, H. Li, J. Yang, W. Liu, J. Lan, Y . Wang, J. Li, J. Wu, Q. Wu et al., “Democratizing{Direct-to-Cell}low earth orbit satellite networks,” in Proceedings of USENIX NSDI, 2024

    2024

  39. [47]

    Dual network computation offloading based on drl for satellite- terrestrial integrated networks,

    D. Li, Y . Sun, J. Peng, S. Cheng, Z. Yin, N. Cheng, J. Liu, Z. Li, and C. Xu, “Dual network computation offloading based on drl for satellite- terrestrial integrated networks,” IEEE TMC, 2024

    2024

  40. [48]

    Sateriot: High-performance ground-space networking for rural iot,

    Y . Ren, A. Gamage, L. Liu, M. Li, S. Chen, Y . Dong, and Z. Cao, “Sateriot: High-performance ground-space networking for rural iot,” in Proceedings of ACM MobiCom, 2024

    2024

  41. [49]

    Co-gps: Energy efficient gps sensing with cloud offloading,

    J. Liu, B. Priyantha, T. Hart, Y . Jin, W. Lee, V . Raghunathan, H. S. Ramos, and Q. Wang, “Co-gps: Energy efficient gps sensing with cloud offloading,” IEEE TMC, 2015

    2015

  42. [50]

    Plug-and-play indoor gps position- ing system with the assistance of optically transparent metasurfaces,

    R. Li, X. Zheng, L. Liu, and H. Ma, “Plug-and-play indoor gps position- ing system with the assistance of optically transparent metasurfaces,” in Proceedings of ACM MobiCom, 2024

    2024

  43. [51]

    Gpsense: Passive sensing with pervasive gps signals,

    H. Dong, M. Cui, N. Wang, L. Qiu, J. Xiong, and W. Wang, “Gpsense: Passive sensing with pervasive gps signals,” in Proceedings of ACM MobiCom, 2024. 10

    2024

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.