Pith. sign in

REVIEW 1 major objections 2 minor 1 cited by

GreenSeg: Ground Segmentation Algorithm for Agricultural Robots in Mediterranean Greenhouses using RGB-D Point Clouds

T0 review · 1 major / 2 minor · reviewed 2026-06-29 · grok-4.3

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

Pith's one-line read GreenSeg applies dual-layer validation on RGB-D point clouds to segment navigable ground in Mediterranean greenhouses with gains up to 19.24% mIoU.

desk verdict GreenSeg improves segmentation metrics in four greenhouse tests but provides no direct evidence on navigation performance. read the letter →

arxiv 2605.25279 v1 pith:56UAKYS7 submitted 2026-05-24 cs.RO

Fernando Cañadas-Aránega , José C. Moreno , José L. Blanco-Claraco This is my paper · ORCID
classification cs.RO
keywords groundsegmentationRGB-Dpointcloudsagriculturalrobotsgreenhousenavigationplanefittingregiongrowingautonomousdepthsensing
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

The paper develops a ground segmentation approach called GreenSeg for robots in Mediterranean greenhouses, where narrow aisles, mixed concrete and soil surfaces, and light reflections from plastic covers create depth sensor errors. It combines robust global plane fitting with a curvature filter to handle terrain changes, then applies seed-point region growing to keep the identified navigable area connected. This setup is tested on an actual robot platform across four different times of day with changing sunlight. If the approach works as described, it would let farms use affordable RGB-D cameras instead of costly LiDAR systems for reliable autonomous movement. The reported results show consistent outperformance over other segmentation techniques, with the largest gains during turns at aisle ends.

What carries the argument

The dual-layer validation strategy that pairs robust global plane fitting and surface curvature filtering with seed-point-based region growing to maintain connected navigable surfaces.

What would settle it

A new test showing segmentation failure or broken spatial continuity when the robot encounters a solar elevation, terrain type, or depth artifact outside the four diurnal conditions used in the AGRICOBIOT I experiments.

Watch

Extended reading notes

Core claim

GreenSeg introduces a dual-layer validation strategy for ground segmentation from RGB-D point clouds: robust global plane fitting combined with a surface curvature filter for terrain adaptability, plus a seed-point-based Region Growing constraint to ensure the spatial continuity of the navigable plane. Experiments on the AGRICOBIOT I platform across four diurnal scenarios demonstrate peak gains of 11.58% in mean Recall and 19.24% in mIoU over benchmark methods, particularly during rotational maneuvers.

Load-bearing premise

The dual-layer validation strategy will maintain performance and spatial continuity across the full range of greenhouse terrains, solar elevations, and sensor artifacts beyond the four tested diurnal scenarios.

Editorial extensions

If this is right

  • The method yields higher mean Recall during critical turns at corridor ends compared with prior segmentation techniques.
  • It produces higher mIoU scores for identifying the navigable plane in heterogeneous greenhouse surfaces.
  • Affordable RGB-D sensors can replace more expensive LiDAR for ground detection in budget-limited facilities.
  • The region-growing step preserves continuity of the traversable area even when terrain changes within an aisle.

Reading between the lines

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

  • The same plane-fitting plus region-growing sequence could be tested on robots operating inside other covered structures that produce similar depth noise.
  • Lower sensor costs from this approach might allow smaller farms to adopt autonomous navigation that was previously uneconomic.
  • Running the algorithm on a different robot chassis or in a greenhouse with different cover materials would check how much the curvature filter depends on the tested conditions.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

1 major / 2 minor

Summary. The manuscript presents GreenSeg, a ground segmentation algorithm for RGB-D point clouds aimed at enabling autonomous navigation for agricultural robots in Mediterranean greenhouses. It proposes a dual-layer validation strategy consisting of robust global plane fitting combined with a surface curvature filter, and a seed-point-based Region Growing constraint. Validation on the AGRICOBIOT I platform across four diurnal scenarios reports peak improvements of 11.58% in mean Recall and 19.24% in mIoU compared to benchmarks, particularly during rotational maneuvers, leading to the conclusion that it enables stable and safe navigation in unstructured, dynamic, budget-constrained environments sensitive to lighting.

Significance. If the reported segmentation improvements hold and translate to navigation performance, the work could offer a practical, low-cost RGB-D based solution for greenhouse automation where LiDAR is economically unfeasible, addressing specific challenges like narrow aisles, heterogeneous terrains, and optical interference from polyethylene covers. The dual-layer approach for terrain adaptability and spatial continuity is a potentially useful contribution to perception for mobile robots in challenging agricultural settings.

major comments (1)
  1. [Abstract] Abstract: The central claim that the algorithm 'enables stable and safe autonomous navigation' in unstructured, dynamic agricultural environments is not supported by the reported evidence. The experiments provide only segmentation metrics (mean Recall and mIoU) from four diurnal scenarios; no navigation-specific results such as path deviation, collision avoidance rates, traversal success, or handling of dynamic obstacles are described. This inference from improved ground segmentation to navigation stability and safety is load-bearing for the paper's primary contribution and requires either additional experimental validation or a more limited claim.
minor comments (2)
  1. [Abstract] The manuscript lacks details on implementation (e.g., specific algorithms for plane fitting, curvature computation, or region growing parameters), error bars or statistical significance of the reported improvements, dataset statistics (number of point clouds, point density), and ablation studies to isolate the contribution of each layer in the dual-layer strategy.
  2. No equations, pseudocode, or parameter values are provided, making reproducibility difficult.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the careful review and the constructive major comment. We agree that the abstract overstates the implications of the segmentation results and will revise the manuscript to adopt a more limited claim supported by the presented evidence.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central claim that the algorithm 'enables stable and safe autonomous navigation' in unstructured, dynamic agricultural environments is not supported by the reported evidence. The experiments provide only segmentation metrics (mean Recall and mIoU) from four diurnal scenarios; no navigation-specific results such as path deviation, collision avoidance rates, traversal success, or handling of dynamic obstacles are described. This inference from improved ground segmentation to navigation stability and safety is load-bearing for the paper's primary contribution and requires either additional experimental validation or a more limited claim.

    Authors: We agree with the referee that the current abstract claim is not directly supported by the experiments, which report only segmentation performance (Recall and mIoU). The manuscript does not include navigation-specific metrics such as path deviation or collision rates. We will revise the abstract (and any corresponding statements in the introduction and conclusion) to remove the assertion that the algorithm 'enables stable and safe autonomous navigation' and instead state that the improved segmentation performance provides a foundation for such navigation in the target environments. No new navigation experiments will be added, as they fall outside the scope of the current study. revision: yes

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: empirical algorithm validated on independent physical experiments

full rationale

The paper proposes GreenSeg as a dual-layer ground segmentation method (global plane fitting + curvature filter + seed-point region growing) and reports its performance via direct comparison against benchmark methods on RGB-D data collected from the AGRICOBIOT I platform across four diurnal scenarios. No equations, fitted parameters, or first-principles derivations are presented that could reduce the reported Recall or mIoU gains to self-referential definitions; the metrics are outcomes of physical testing rather than statistical fits or renamed inputs. No self-citations are invoked as load-bearing uniqueness theorems, and the navigation-enabling claim is an interpretive inference from the segmentation results rather than a tautological redefinition of those results. The derivation chain is therefore self-contained against external benchmarks.

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

Based solely on the abstract; no explicit free parameters, axioms, or invented entities are described. The method appears to rely on standard point-cloud primitives without introducing new postulated entities.

how reviews work

0 comments
Cite this review

Pith. "Pith review of GreenSeg: Ground Segmentation Algorithm for Agricultural Robots in Mediterranean Greenhouses using RGB-D Point Clouds." pith.science (2026). https://pith.science/paper/56UAKYS7

@misc{pith2026260525279,
  author       = {Pith},
  title        = {Pith review of: GreenSeg: Ground Segmentation Algorithm for Agricultural Robots in Mediterranean Greenhouses using RGB-D Point Clouds},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/56UAKYS7}},
  note         = {Machine review of arXiv:2605.25279}
}
read the original abstract

Greenhouse agriculture in the Mediterranean region faces significant automation challenges due to its unique structural and environmental constraints. These environments are characterized by extremely narrow aisles, heterogeneous terrains ranging from concrete to tilled soil and severe optical interference caused by polyethylene covers, which induce specular reflections and "ghost points" in depth sensors. While autonomous navigation is essential for digitizing agricultural tasks, traditional solutions often rely on expensive 3D LiDAR systems that are economically unscalable for most facilities. To address this, this paper presents GreenSeg, a robust perception framework for autonomous navigation using RGB-D sensing. The proposed method introduces a dual-layer validation strategy: a robust global plane fitting combined with a surface curvature filter for terrain adaptability, and a seed-point-based Region Growing constraint to ensure the spatial continuity of the navigable plane. Experimental validation was conducted using the AGRICOBIOT I platform across four diurnal scenarios with varying solar elevations. The results show that GreenSeg consistently outperforms benchmark segmentation methods, achieving peak improvements of 11.58% in mean Recall and 19.24% in mIoU during critical rotational maneuvers at the end of corridors. These findings confirm that the proposed algorithm enables stable and safe autonomous navigation in unstructured, dynamic agricultural environments that are subject to budget constraints and sensitive to lighting conditions.

Figures

Figures reproduced from arXiv: 2605.25279 by the authors.

Figure 1
Figure 1. AgroConnect greenhouse 3.1.2. AGRICOBIOT I robot The experimental validation involves the deployment of the AGRICO￾BIOT I (López-Gázquez et al., 2023) mobile platform, seamlessly integrated into a fog-based architecture to facilitate real-time data exchange across the facility’s broader operational network. Built upon a commercial Husky chas￾sis by Clearpath Robotics® , this platform employs a differential drive sys… view at source ↗
Figure 2
Figure 2. shows various views of the AGRICOBIOT I robot. (a) AGRICOBIOT I in a corridor with high plants(b) AGRICOBIOT I in a corridor with low plants [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 3
Figure 3. Example of how to use the framework with a jet-type color intensity Intel [PITH_FULL_IMAGE:figures/full_fig_p015_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Camera location in robot navigation in all possible scenarios a robot may encounter in a greenhouse, which will be analyzed and studied in this section. The output of the base algorithm undergoes a second phase for soil classification, where Pground and Pobs are evalua…
Figure 5
Figure 5. Figure 5: Comparison of algorithms 4.3. Analysis of the most adverse conditions in the Mediterranean Green￾house The performance of the optical system is intrinsically linked to the ge￾ometric orientation of the incident light, defined by the sun’s elevation and azimuth angles J…
Figure 6
Figure 6. Figure 6: Example of solar elevation and azimuth in the greenhouse [PITH_FULL_IMAGE:figures/full_fig_p021_6.png]
Figure 7
Figure 7. Figure 7: Tests paths. The red line represents the path moving through the central [PITH_FULL_IMAGE:figures/full_fig_p025_7.png]
Figure 8
Figure 8. Figure 8: Central corridor test at four solar conditions. Green points indicates the ground, [PITH_FULL_IMAGE:figures/full_fig_p027_8.png]
Figure 9
Figure 9. Figure 9: Crop-row traversal test at four solar conditions. Green points indicates the [PITH_FULL_IMAGE:figures/full_fig_p029_9.png]
Figure 10
Figure 10. Figure 10: End-of-corridor turning test at four solar conditions. Green points indicates [PITH_FULL_IMAGE:figures/full_fig_p030_10.png]
Figure 11
Figure 11. Figure 11: Changing to the next corridor test at four solar conditions. Green points [PITH_FULL_IMAGE:figures/full_fig_p032_11.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Integrated cloud-based architecture for robot-robot and human-robot collaboration using ROS 2--MQTT in Mediterranean Greenhouses

    cs.RO 2026-06 unverdicted novelty 3.0 of 10

    The paper introduces and tests a hybrid architecture integrating ROS 2 with MQTT and the FIWARE platform for multi-agent robotic collaboration in greenhouses with communication constraints.

Reference graph

Works this paper leans on

9 extracted references · 5 canonical work pages · cited by 1 Pith paper

  1. [1]

    A PID- based control architecture for mobile robot path planning in greenhouses

    36 Cañadas-Aránega, F., Moreno, J.C., Blanco-Claraco, J.L., 2024b. A PID- based control architecture for mobile robot path planning in greenhouses. IFAC-PapersOnLine 58, 503–508. doi:https://doi.org/10.1016/j.if acol.2024.08.112. Cañadas-Aránega, F., Moreno, J.C., Blanco-Claraco, J.L., Giménez, A., Ro- dríguez, F., Sánchez-Hermosilla, J., 2024c. Autonomou...

    work page doi:10.1016/j.if 2024

  2. [2]

    https://doi.org/10.48550/arXiv

    A ros2 benchmarking framework for hierar- chical control strategies in mobile robots for mediterranean greenhouses. arXiv preprint arXiv:2602.15162 doi:https://doi.org/10.48550/arXiv .2602.15162. De Silva, R., Cielniak, G., Wang, G., Gao, J.,

    work page Pith review doi:10.48550/arxiv

  3. [3]

    Agricultural Robotics: The Future of Robotic Agriculture

    Agricultural robotics: thefutureofroboticagriculture. arXivpreprintarXiv:1806.06762 . Firkat, E., An, F., Peng, B., Zhang, J., Mijit, T., Ahat, A., Zhu, J., Ham- dulla, A.,

    work page Pith review arXiv

  4. [4]

    Open Edge Platform Documentation

    3D Pointcloud Groundfloor Segmentation for Re- alSense™Camera and 3D LiDAR. Open Edge Platform Documentation. URL:https://docs.openedgeplatform.intel.com/dev/edge-ai-s uites/robotics-ai-suite/robotics/dev_guide/tutorials_amr/pe rception/pointcloud-groundfloor-segmentation.html. accessed: 2026-04-30. Jafarbiglu, H., Pourreza, A.,

    2026

  5. [5]

    arXiv preprint arXiv:2603.27344

    Ter- raseg: Self-supervised ground segmentation for any lidar. arXiv preprint arXiv:2603.27344 . López-Gázquez, A., Mañas-Alvarez, F., Moreno, J., Cañadas-Aránega, F., Sánchez, J.,

    work page arXiv

  6. [6]

    The marathon 2: A navigation system, in: 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), IEEE, Las Vegas, NV, USA. pp. 2718–2725. doi:10.1109/IROS45743.2020.9341207. Mijit, T., Firkat, E., Yuan, X., Liang, Y., Zhu, J., Hamdulla, A.,

    work page doi:10.1109/iros45743.2020.9341207 2020

  7. [7]

    Real-time semantic segmentation of crop and weed for precision agriculture robots leveraging background knowledge in cnns, in: 2018 IEEE international conference on robotics and automation (ICRA), IEEE. pp. 2229–2235. Moreno, J., Rodríguez, F., Sánchez-Hermosilla, J., Giménez, A., Sánchez- Molina, J.,

    2018

  8. [8]

    Gnd- net: Fast ground plane estimation and point cloud segmentation for au- tonomous vehicles, in: 2020 IEEE/RSJ international conference on intel- ligent robots and systems (IROS), IEEE. pp. 2150–2156. 39 Qian, X., Ye, C.,

    2020

Show all 9 references
  1. [9]

    Atmospheric Measurement Techniques 13, 1953–1961

    Segcloud: A novel cloud image segmentation model using a deep convolutional neural network for ground-based all- sky-view camera observation. Atmospheric Measurement Techniques 13, 1953–1961. Xu, Y., Ghamisi, P.,

    1953

Pith tools

Reviewed June 29, 2026 · model on record in the stance chip above.