Pith. sign in

REVIEW 4 major objections 5 minor 89 references

DeepForest: Sensing Into Self-Occluding Volumes of Vegetation With Aerial Imaging

T0 review · 4 major / 5 minor · reviewed 2026-08-09 · deepseek-v4-flash

Pith's one-line read The paper claims that ordinary aerial multispectral images, fused by synthetic-aperture focal stacking and cleaned by pre-trained 3D CNNs, can expose volumetric reflectance and per-voxel NDVI through self-occluding forest canopies.

desk verdict A solid, honestly-scoped method paper that demonstrates in simulation a cheap passive route to volumetric vegetation reflectance; the real-world deep-canopy claim remains unproven. read the letter →

arxiv 2502.02171 v4 pith:IWYCTEOP submitted 2025-02-04 cs.CV eess.IV

classification cs.CVeess.IV
keywords syntheticapertureimagingmultispectralaerial3DconvolutionalneuralnetworkforeststructurevolumetricreflectanceNDVIstackocclusionremovaldroneremotesensing
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 seeks to establish that conventional aerial multispectral imaging, not just LiDAR or radar, can sense deep into self-occluding vegetation volumes such as forests. The method builds synthetic-aperture focal stacks from many drone images, then uses pre-trained 3D convolutional networks, one per depth layer, to subtract the blur contributed by out-of-focus branches and leaves. Against simulated ground truth in procedural broadleaf forests, the correction yields roughly sevenfold average error reduction (from about twofold to twelvefold) across 220 to 1680 trees per hectare. In a field experiment, corrected red and near-infrared channels produced an NDVI stack whose top layer matched photogrammetrically reconstructed camera NDVI with mean squared error 0.05 after sensor mapping. If correct, the approach would let established drone and aircraft camera platforms produce volumetric vegetation indices throughout the canopy, not just at its surface.

What carries the argument

The central object is the synthetic-aperture focal stack together with its asymmetric inverted-pyramid receptive field: for a point at focal distance $f$, every out-of-focus occluder inside the frustum spanned by the synthetic aperture contributes a spread signal to that point's value. The correction machinery is a per-layer 3D CNN that maps a downsampled patch tensor from this receptive field to a corrected reflectance value; the paper's finding is that the redundancy of focal stacks allows extreme downsampling, so 2x2x20 patches and a network of roughly 6.1 million parameters per layer suffice. One network is trained for each of 440 depth layers, the same weights are reused across spectral bands, and training one layer takes about 15 minutes on a single GPU.

What would settle it

Fly the same 24 m x 24 m, 9x9-pose synthetic-aperture scan over a real forest while independently measuring the below-canopy vegetation with terrestrial LiDAR or destructive sampling, then compare the corrected reflectance and NDVI stacks voxel-by-voxel against those measurements. If the below-canopy error is no better than the uncorrected focal stack, the simulation-trained correction has not transferred to reality.

Watch

Extended reading notes

Core claim

The central claim is that the out-of-focus blur contaminating each layer of a synthetic-aperture focal stack is a learnable function of the depth layer and the local occlusion pattern, and that a 3D CNN trained on procedural forests can suppress it well enough to recover low-frequency volumetric reflectance. Classical 3D deconvolution cannot do this job because the occluders are opaque and the receptive field is shift-variant, so the paper replaces deconvolution with learned per-layer correction. Each network sees only a heavily downsampled patch tensor sampled from the inverted-pyramid receptive field, and a patch size of 2x2x20 suffices, making the models small enough to train per layer. The networks are trained on white-light simulation and applied across spectral channels on the assumption that macroscopic defocus is wavelength-invariant. After correction, per-layer error stays roughly constant with depth even though uncorrected error grows with accumulated occlusion, and the output is a low-frequency reflectance stack rather than a sharp voxel segmentation.

Load-bearing premise

The load-bearing premise is that the computer-generated broadleaf forests block light the same way real forests do, because the correction network is trained only on simulated trees and the field experiment never measures what is actually below the canopy.

Editorial extensions

If this is right

  • Vegetation indices such as NDVI become computable per voxel from passive multispectral camera data, so health and biomass estimates can be stratified by canopy depth instead of being limited to the visible top layer.
  • The approach slots onto existing drone platforms: a 30 m x 30 m plot scanned from 9x9 poses inside a 24 m x 24 m synthetic aperture at 35 m altitude produced a 440x440x440 volume from an 18-minute flight.
  • Because each depth layer is corrected by its own network, the method is most useful exactly where occlusion is worst: uncorrected error grows toward the ground, while corrected error stays flat, yielding the largest relative gains (~7x average, up to ~12x) in deep layers.
  • Retraining is required when forest type, season, or aperture geometry changes, but the paper reports the cost is modest, about 15 minutes per depth layer on a single GPU.
  • With sensor mapping, corrected red and near-infrared stacks reproduce top-layer NDVI from the original camera image with MSE 0.05, the paper's only quantitative check on real data.

Reading between the lines

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

  • An untested extension implied by the paper is validation against independent below-canopy measurements, such as terrestrial laser scans or understory censuses; the field experiment compares only the top vegetation layer, so the deep-layer improvement is demonstrated in simulation but not yet in the real world.
  • The 2x2x20 patch finding suggests a broader principle: strongly occluded focal stacks carry enough redundant angular information that extremely sparse receptive-field sampling works, which may transfer to other self-occluding volume-recovery problems.
  • The paper leaves depth-based void filtering as future work; if void points were identified, the low-frequency reflectance stacks would sharpen toward camera-limited spatial detail, potentially enabling volumetric leaf-area-density (PAD) profiles.
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, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes DeepForest, a method for converting aerial multispectral focal stacks acquired by synthetic-aperture imaging with drones into volumetric reflectance stacks. A 3D CNN is trained on procedural forest simulations to suppress out-of-focus contributions from occluders, and the corrected stacks are combined into vegetation-index stacks such as NDVI. Simulation results show an average ~x7 improvement (min ~x2, max ~x12) in MSE against simulated ground truth for forest densities of 220–1680 trees/ha. A field experiment on a mixed broadleaf forest reports an MSE of 0.05 for top-layer NDVI after a linear sensor-mapping step.

Significance. If the deep-layer recovery were validated against independent below-canopy ground truth, this would be a significant contribution to remote sensing, because it would enable passive optical cameras to provide volumetric vegetation information at lower cost and higher spectral resolution than active LiDAR or radar systems. The paper is notable for its open data and code, its explicit discussion of void points and training limitations, and its honest statement that models must be retrained for different vegetation types. These strengths make the simulation-level results credible as an in-distribution proof of concept, but the central real-world claim is not yet supported by the field experiment.

major comments (4)
  1. [Results, Field Experiment (Fig. 9A)] The only quantitative field validation is against the top vegetation layer: the paper states that 'measured ground truth data for deeper vegetation is unavailable.' Because the abstract and title claim sensing 'deep into self-occluding vegetation volumes, such as forests,' the field experiment does not test the deep-layer prediction. A revision should add an independent below-canopy reference (e.g., UAV or terrestrial LiDAR of the same plot, or leaf-off photography) or explicitly limit the real-world claim to the top layer.
  2. [Results, Eq. (2) sensor mapping] The sensor-mapping step rescales each corrected reflectance stack using the mean and standard deviation of the photogrammetrically reconstructed top-layer points matched to the center camera image. The reported top-layer MSE of 0.05 in Fig. 9A is therefore partly enforced by construction and is not an independent validation of the CNN correction. The paper should report the top-layer MSE of the corrected stack before sensor mapping and clarify what exactly the field experiment validates.
  3. [Training, Validation, and Inference; Fig. 7; Generalization] The x7 average improvement (min x2, max x12) is computed against ground truth generated by the same procedural forest simulator used to create the training patches; this measures in-distribution interpolation, not transfer to real forest occlusion statistics. The Discussion's Generalization section concedes that 'our models must be retrained with adapted procedural forest parameters' for other vegetation types. The abstract's broad claim that the approach 'allows sensing deep into self-occluding vegetation volumes, such as forests' overstates the evidence. Add cross-domain tests (e.g., train on one procedural parameter set and test on another, or simulate a field plot from LiDAR data) or constrain the claim in the abstract.
  4. [Reconstruction Error Suppression; Limitations] The paper acknowledges that void points cannot be learned and that the model 'approximates low (yet not necessarily zero) reflectance values that are noisy in the lateral and axial directions.' During inference all points are processed, and the field NDVI stack is thresholded only by NDVI >= 0.33. The reported 32.75% 'biomass' estimate therefore includes void regions and is not a validated volumetric measure. The paper should add a void-confidence channel or explicitly state that quantitative ecological estimates are not yet reliable until void point classification is solved.
minor comments (5)
  1. [Eq. (1)] The notation in Eq. (1) is unclear: the position of h in the denominator and the structure of the point-spread expression should be clarified, and the integration bounds should be defined more explicitly.
  2. [Table 1] The text refers to 'the improvement factor (last column),' but the printed Table 1 does not contain an improvement-factor column; the table should be completed or the text corrected.
  3. [Introduction] The sentence 'while SAR does not support multi-spectral measurement' begins with a lowercase 'while' after a period; capitalize and fix the punctuation.
  4. [Data S10] Data S10 refers to 'COMAP' where 'COLMAP' is meant; correct the typo.
  5. [Supplementary Data S1] The folder name 'integrals_full_respolution' contains a typo; it should be 'integrals_full_resolution'.

Circularity Check

2 steps flagged · score 6.0 of 10

Field top-layer validation is partly enforced by the sensor-mapping calibration, and the simulated x7 gain is an in-simulator supervised-fit metric; the deep-sensing claim is supported by interpolation, not independent prediction.

  1. fitted input called prediction [Vegetation-Index Estimation from Field Experiments, Eq. 2 and Fig. 9A]
    "Since these points are unoccluded, their reflectance values should match the reflectance of the unoccluded points visible in the original camera images. We approximate this by matching their reflectance statistics and by correcting the reflectance stacks: R'=σC(R-μR)/σR + μC ... Here, we achieved an overall MSE of 0.05 ... with sensor mapping (MSE=0.2, RMSE%=22.4% without sensor mapping). This comparison is conducted as an initial step to assess our approach effectiveness in correcting the reflectance values for the top vegetation."

    Equation 2 is an affine map whose parameters (μR, σR) are computed from the corrected reflectance-stack values at the very top-layer point cloud that is later compared, and (μC, σC) from the center camera image. After mapping, the mean and standard deviation of the top-layer reflectance values in each channel equal the camera image's statistics by construction. The reported top-layer NDVI MSE of 0.05 is therefore not an independent test of the learned 3D-CNN correction; it is largely a check of the affine sensor-mapping calibration applied to the same points used to fit it.

  2. fitted input called prediction [Abstract and Reconstruction Error Suppression / Simulated Results]
    "The training data which provides ground-truth values comes from simulated procedural forests with varying vegetation parameters ... Compared with simulated ground truth, our correction leads to ~x7 average improvements (min: ~x2, max: ~x12) for forest densities of 220 trees/ha - 1680 trees/ha."

    The CNN is trained with MSE loss against ground-truth reflectance slices extracted from the GAZEBO procedural forest simulator (near/far clipping), and the headline improvement is measured against simulated ground truth generated by that same simulator. Thus the x7 figure quantifies how well the fitted network reproduces the simulator's own reflectance distribution on held-out instances of that distribution; it is a supervised-fit quality metric, not an independent prediction for real forests. The paper's Generalization section concedes that models must be retrained for other occlusion statistics, confirming that the result is tied to the training simulator.

full rationale

The paper's core method is transparent: focal stacks are computed by synthetic-aperture imaging, and a 3D CNN trained on procedural forests suppresses out-of-focus occluder signal. The simulated evaluation is in-distribution: both training targets and test ground truth come from the same GAZEBO procedural simulator, so the x7 improvement demonstrates interpolation within that simulator rather than transfer to real below-canopy reflectance. The field experiment provides no below-canopy ground truth, and the reported top-layer MSE of 0.05 is weakened by the sensor-mapping step (Eq. 2), which forces the mapped top-layer reflectance statistics to match the center camera image and is calibrated on the same point cloud used for the comparison. No load-bearing self-citation chain or uniqueness argument is present; the cited AOS prior work supplies the focal-stack computation but is not used to forbid alternatives. The central deep-sensing claim therefore rests partly on a fitted, self-consistent simulation and a partially construction-enforced field metric, warranting a partial circularity score rather than a fully independent validation.

Assumptions & free parameters 4 free parameters · 6 assumptions · 0 invented entities

The method's central claim rests on a learned mapping trained entirely in simulation. The main free parameters are the CNN weights and the hand-chosen receptive-field sampling; the key axioms are that Eq. 1 models the imaging process, that defocus is wavelength-invariant, and that procedural forests reproduce real occlusion statistics. No new physical entities are introduced.

free parameters (4)
  • 3D CNN weights = 6,143,009 per layer; 2,703,923,960 total for 440 layers
    Learned by MSE minimization on simulated procedural forest focal stacks; the core correction is a fitted neural model, not a derived inversion.
  • Receptive-field patch resolution (Pw x Ph x Pd) = 2x2x20
    Chosen from Table 1 as a balance between model size and convergence; this sampling resolution defines the information available to every corrected voxel.
  • Volume resolution (Vw x Vh x Vd) = 440x440x440 voxels
    Selected to match 440x440 px images over a 20 m height; affects voxel size and network input structure.
  • Sensor-mapping moments (muC, sigmaC, muR, sigmaR) = Per channel from field imagery
    Eq. 2 aligns corrected stack reflectance to the center camera image using top-layer statistics; used before computing NDVI and before reporting the top-layer field MSE.
assumptions (6)
  • domain assumption The synthetic-aperture integral model in Eq. 1 approximates the actual imaging of opaque vegetation volumes.
    The entire focal-stack construction and correction framework assumes this model of signal integration over the synthetic aperture.
  • domain assumption At macroscopic scale, defocus blur is invariant to wavelength, so white-light-trained networks can be applied to each multispectral band.
    Stated in Materials and Methods and in Discussion; without it, per-band retraining would be needed.
  • domain assumption Procedural GAZEBO forests with the stated tree parameters reproduce the occlusion statistics of real European broadleaf forests.
    Training data comes entirely from this simulator; the paper acknowledges retraining is needed for other vegetation types.
  • domain assumption Averaged RGB renderings of thin depth slices provide ground-truth reflectance for training.
    Ground truth is defined by the simulator's near and far clipping planes and channel averaging; this is not validated against measured reflectance.
  • domain assumption Photogrammetric COLMAP reconstruction of the top canopy layer is accurate enough to serve as reference for sensor mapping and field validation.
    Used in Eq. 2 and Fig. 9A; errors in the COLMAP point cloud would propagate into the sensor-mapping statistics and the reported MSE.
  • domain assumption The training strategy using only non-void points is valid for inference on all voxels.
    Training converges only when void points are excluded, but inference cannot distinguish void from non-void points, so the learned mapping is applied everywhere.

how reviews work

0 comments
Cite this review

Pith. "Pith review of DeepForest: Sensing Into Self-Occluding Volumes of Vegetation With Aerial Imaging." pith.science (2026). https://pith.science/paper/IWYCTEOP

@misc{pith2026250202171,
  author       = {Pith},
  title        = {Pith review of: DeepForest: Sensing Into Self-Occluding Volumes of Vegetation With Aerial Imaging},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IWYCTEOP}},
  note         = {Machine review of arXiv:2502.02171}
}
read the original abstract

Access to below-canopy volumetric vegetation data is crucial for understanding ecosystem dynamics. We address the long-standing limitation of remote sensing to penetrate deep into dense canopy layers. LiDAR and radar are currently considered the primary options for measuring 3D vegetation structures, while cameras can only extract the reflectance and depth of top layers. Using conventional, high-resolution aerial images, our approach allows sensing deep into self-occluding vegetation volumes, such as forests. It is similar in spirit to the imaging process of wide-field microscopy, but can handle much larger scales and strong occlusion. We scan focal stacks by synthetic-aperture imaging with drones and reduce out-of-focus signal contributions using pre-trained 3D convolutional neural networks with mean squared error (MSE) as the loss function. The resulting volumetric reflectance stacks contain low-frequency representations of the vegetation volume. Combining multiple reflectance stacks from various spectral channels provides insights into plant health, growth, and environmental conditions throughout the entire vegetation volume. Compared with simulated ground truth, our correction leads to ~x7 average improvements (min: ~x2, max: ~x12) for forest densities of 220 trees/ha - 1680 trees/ha. In our field experiment, we achieved an MSE of 0.05 when comparing with the top-vegetation layer that was measured with classical multispectral aerial imaging.

Figures

Figures reproduced from arXiv: 2502.02171 by the authors.

Figure 1
Figure 1. Photogrammetry vs. synthetic-aperture imaging in the presence of partial occlusion. Dense forest vegetation captured with multiple aerial images at altitude h within sampling area a. The reconstruction of a scene point (red) at distance d=f (depth=focal distance) is occluded in partial views by another scene point (blue). In photogrammetry (A), incorrectly matched features lead to reconstruction errors. In synthetic… view at source ↗
Figure 2
Figure 2. Synthetic-aperture imaging. Multiple spectral bands (green, near-infrared, red, and red edge) captured at an altitude of 35 m above ground level (AGL) of a 30 m x 30 m mixed forest plot (top 4 rows). White-light procedural forest simulation under the same sampling conditions (bottom row). Conventional narrow-aperture camera images (left) and wide-aperture integral images (right) resulting from synthetic￾aperture ima… view at source ↗
Figure 3
Figure 3. Reconstruction architecture. Multi-view multispectral (SC=spectral channel) aerial images are converted into multispectral focal stacks by synthetic-aperture imaging. Out-of-focus errors in reflectance values of the focal stacks are suppressed using 3D convolutional neural networks (CNNs) that are pre-trained individually for each focal-stack layer, which yields multispectral reflectance stacks. The reflectance stac… view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Sampling the receptive field. The receptive field includes all out-of-focus points that contribute blur to specific in-focus points (red). These out-of-focus regions lie within the frustum that spans from the in-focus points to the outer extent of the synthetic apertur…
Figure 5
Figure 5. Figure 5: 3D CNN model. Our 3D convolutional neural network model takes receptive field patches as input and outputs corrected reflectance values. It consists of eight convolutional layers of an increasing number of activation slices (32..256) with the same resolutions as the re…
Figure 6
Figure 6. Figure 6: Simulated reflectance correction. Different perspectives of a simulated sparse procedural forest plot (top row). Uncorrected focal stack (second row) and corrected reflectance stack (third row). Photometrically reconstructed (geometry and reflectance) point cloud (bott…
Figure 7
Figure 7. Figure 7: Error for various vegetation layers and forest densities. (A) Development of the mean squared error (MSE) over various vegetation layers (0 m AGL = ground, 20 m AGL = highest tree top, simulated density: 220 trees/ha). The MSE of uncorrected reflectance values increase…
Figure 8
Figure 8. Figure 8: Reflectance correction and sensor mapping. Sensor mapping by matching reflectance statistics of the top vegetation layer’s reconstructed point cloud with reflectance statistics of original camera image at the SA’s center (A). The red squares indicate the same field of …
Figure 9
Figure 9. Figure 9: NDVI stack. NDVI stack values determined from extracted top vegetation-layer point cloud (result of merging the photogrammetric reconstructions of the NIR and RED channels) with and without sensor mapping of the two spectral channels vs. NDVI image computed from origin…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

89 extracted references · 78 canonical work pages

  1. [1]

    Combining Environmental, Multispectral, and LiDAR Data Improves Forest Type Classification: A Case Study on Mapping Cool Temperate Rainforests and Mixed Forests

    Trouvé R, Jiang R, Fedrigo M, White MD, Kasel S, Baker PJ, Nitschke CR. Combining Environmental, Multispectral, and LiDAR Data Improves Forest Type Classification: A Case Study on Mapping Cool Temperate Rainforests and Mixed Forests. Remote Sens. 2023;15:60

  2. [2]

    Towards accurate individual tree parameters estimation in dense forest: optimized coarse-to-fine algorithms for registering UAV and terrestrial LiDAR data

    Zhao Y, Im J, Zhen Z, Zhao Y. Towards accurate individual tree parameters estimation in dense forest: optimized coarse-to-fine algorithms for registering UAV and terrestrial LiDAR data. GIScience & Remote Sensing. 2023;60:2197281

  3. [3]

    Forest Emissions Reduction Assessment Using Optical Satellite Imagery and Space LiDAR Fusion for Carbon Stock Estimation

    Jiao Y, Wang D, Yao X, Wang S, Chi T, Meng Y. Forest Emissions Reduction Assessment Using Optical Satellite Imagery and Space LiDAR Fusion for Carbon Stock Estimation. Remote Sens. 2023;15(5):1410

  4. [4]

    Biomass Estimation of Subtropical Arboreal Forest at Single Tree Scale Based on Feature Fusion of Airborne LiDAR Data and Aerial Images

    Yan M, Xia Y, Yang X, Wu X, Yang M, Wang C, Hou Y, Wang D. Biomass Estimation of Subtropical Arboreal Forest at Single Tree Scale Based on Feature Fusion of Airborne LiDAR Data and Aerial Images. Sustainability. 2023;15(2):1676

  5. [5]

    Crop Type Classification by DESIS Hyperspectral Imagery and Machine Learning Algorithms

    Farmonov N, Amankulova K, Szatmári J, Sharifi A, Abbasi-Moghadam D, Nejad SMM, Mucsi L. Crop Type Classification by DESIS Hyperspectral Imagery and Machine Learning Algorithms. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing. 2023;16:1576-1588

  6. [6]

    Hong SJ, Park S, Lee A, Kim SY, Kim E, Lee CH. Kim G. Nondestructive prediction of pepper seed viability using single and fusion information of hyperspectral and X-ray images. Sensors and Actuators A: Physical. 2023;350:114151. Page 23 of 34

  7. [7]

    Hyperspectral remote sensing to assess weed competitiveness in maize farmland ecosystems

    Lou Z, Quan L, Sun D, Li H, Xia F. Hyperspectral remote sensing to assess weed competitiveness in maize farmland ecosystems. Science of The Total Environment. 2022;844:157071

  8. [8]

    Early Detection of Dendroctonus valens Infestation at Tree Level with a Hyperspectral UAV Image

    Gao B, Yu L, Ren L, Zhan Z, Luo Y. Early Detection of Dendroctonus valens Infestation at Tree Level with a Hyperspectral UAV Image. Remote Sens. 2023;15:407

Show all 89 references
  1. [9]

    Combining novel feature selection strategy and hyperspectral vegetation indices to predict crop yield

    Fei S, Li L, Han Z, Chen Z, Xiao Y. Combining novel feature selection strategy and hyperspectral vegetation indices to predict crop yield. Plant Methods. 2022;18:119

  2. [10]

    Real-time defect inspection of green coffee beans using NIR snapshot hyperspectral imaging

    Chen SY, Chiu MF, Zou XW. Real-time defect inspection of green coffee beans using NIR snapshot hyperspectral imaging. Computers and Electronics in Agriculture. 2022;197: 106970

  3. [11]

    Detecting Asymptomatic Infections of Rice Bacterial Leaf Blight Using Hyperspectral Imaging and 3-Dimensional Convolutional Neural Network With Spectral Dilated Convolution

    Cao Y, Yuan P, Xu H, Martínez-Ortega JF, Feng J, Zhai Z. Detecting Asymptomatic Infections of Rice Bacterial Leaf Blight Using Hyperspectral Imaging and 3-Dimensional Convolutional Neural Network With Spectral Dilated Convolution. Front. Plant Sci. 2022;13: 963170

  4. [12]

    Remote Estimation of Chlorophyll Content in Higher Plant Leaves

    Gitelson AA, Merzlyak MN. Remote Estimation of Chlorophyll Content in Higher Plant Leaves. Int. J. Remote Sens. 1997;18:2691–2697

  5. [13]

    Canopy Top, Height and Photosynthetic Pigment Estimation Using Parrot Sequoia Multispectral Imagery and the Unmanned Aerial Vehicle (UAV)

    Kopačková-Strnadová V, Koucká L, Jelének J, Lhotáková Z, Oulehle F. Canopy Top, Height and Photosynthetic Pigment Estimation Using Parrot Sequoia Multispectral Imagery and the Unmanned Aerial Vehicle (UAV). Remote Sens. 2021;13:705

  6. [14]

    Construction of 3D maps of vegetation indices retrieved from UAV multispectral imagery in forested areas

    Villacrés J, Cheein FAA. Construction of 3D maps of vegetation indices retrieved from UAV multispectral imagery in forested areas. Biosystems Engineering.2022;213:76

  7. [15]

    In-situ fruit analysis by means of LiDAR 3D point cloud of normalized difference vegetation index (NDVI)

    Tsoulias N, Saha KK, Zude-Sasse M. In-situ fruit analysis by means of LiDAR 3D point cloud of normalized difference vegetation index (NDVI). Computers and Electronics in Agriculture. 2023;205:107611

  8. [16]

    Evaluating the Combined Use of the NDVI and High-Density Lidar Data to Assess the Natural Regeneration of P

    Míguez C, Fernández C. Evaluating the Combined Use of the NDVI and High-Density Lidar Data to Assess the Natural Regeneration of P. pinaster after a High-Severity Fire in NW Spain. Remote Sens. 2023;15:1634

  9. [17]

    Integrating UAV LiDAR and multispectral data to assess forest status and map disturbance severity in a West African forest patch

    Iheaturu CJ, Hepner S, Batchelor JL, Agonvonon GA, Akinyemi FO, Wingate VR, Speranza CI. Integrating UAV LiDAR and multispectral data to assess forest status and map disturbance severity in a West African forest patch. Ecological Informatics. 2024; 84:102876

  10. [18]

    L-Band Synthetic Aperture Radar and Its Application for Forest Parameter Estimation, 1972 to 2024: A Review

    Ye Z, Long J, Zhang T, Lin B, Lin H. L-Band Synthetic Aperture Radar and Its Application for Forest Parameter Estimation, 1972 to 2024: A Review. Plants. 2024; 13:2511

  11. [19]

    Measurements of Forest Biomass Change Using P-Band Synthetic Aperture Radar Backscatter

    Sandberg G, Ulander LMH, Wallerman J, Fransson JES. Measurements of Forest Biomass Change Using P-Band Synthetic Aperture Radar Backscatter. In IEEE Geosci. Remote Sens. 2014;52:6047-6061. Page 24 of 34

  12. [20]

    P-band SAR for ground deformation surveying: Advantages and challenges

    Xu Y, Lu Z, Bürgmann R, Hensley S, Fielding E, Kim J. P-band SAR for ground deformation surveying: Advantages and challenges. Remote Sens. Environ. 2023; 287:113474

  13. [21]

    Virtual constellations for global terrestrial monitoring

    Wulder MA, Hilker T, White JC, Coops NC, Masek JG, Pflugmacher D, Crevier Y. Virtual constellations for global terrestrial monitoring. Remote Sens. Environ. 2015; 170:62-76

  14. [22]

    Radar vegetation indices for monitoring surface vegetation: Developments, challenges, and trends

    Hu X, Li L, Huang J, Zeng Y, Zhang S, Su Y, Hong Y, Hong Z. Radar vegetation indices for monitoring surface vegetation: Developments, challenges, and trends. Science of The Total Environment. 2024;945:173974

  15. [23]

    Combining lidar and synthetic aperture radar data to estimate forest biomass: status and prospects

    Kaasalainen S, Holopainen M, Karjalainen M, Vastaranta M, Kankare V, Karila K, Osmanoglu B. Combining lidar and synthetic aperture radar data to estimate forest biomass: status and prospects. Forests. 2015;6(1):252-270

  16. [24]

    Structure-from-Motion Revisited

    Schönberger JL, Frahm JM. Structure-from-Motion Revisited. Paper presented at: CVPR

  17. [25]

    Pixelwise view selection for unstructured multi-view stereo

    Schönberger JL, Zheng E, Frahm JM, Pollefeys M. Pixelwise view selection for unstructured multi-view stereo. Paper presented at: ECCV 2016. Proceeding of European Conference on Computer Vision;2016 Oct11-14; Amsterdam, The Netherlands

  18. [26]

    Global structure-from-motion revisited

    Pan L, Baráth D, Pollefeys M, Schönberger JL. Global structure-from-motion revisited. Paper presented at: ECCV 2024. Proceeding of European Conference on Computer Vision; 2024 Sep 29 – Oct 4; Milan, Italy

  19. [27]

    Atutorial on synthetic aperture radar

    Moreira A, Prats-Iraola P, Younis M, Krieger G, Hajnsek I, Papathanassiou KP. Atutorial on synthetic aperture radar. IEEE Geosci. Remote Sens. 2013;1:6–43

  20. [28]

    Synthetic aperture radar imaging using a small consumer drone

    Li CJ, Ling H. Synthetic aperture radar imaging using a small consumer drone. Proceeding IEEE International Symposium on Antennas and Propagation USNC/URSI National Radio Science Meeting;2015 Jul 19; Vancouver, BC, Canada

  21. [29]

    Synthetic aperture radar interferometry

    Rosen PA, Hensley S, Joughin IR, Li FK, Madsen SN, Rodriguez E, Goldstein RM. Synthetic aperture radar interferometry. Proc. IEEE. 2000;88:333–382

  22. [30]

    Interferometric synthetic aperture microscopy (ISAM)

    Ralston TS, Marks DL, Carney PS, Boppart SA. Interferometric synthetic aperture microscopy (ISAM). Nat. Phys. 2007;3:965–1004

  23. [31]

    Synthetic aperture sonar: a review of current status

    Hayes MP, Gough PT. Synthetic aperture sonar: a review of current status. IEEE J. Ocean. Eng. 2009;34:207–224

  24. [32]

    Synthetic aperture ultrasound imaging

    Jensen JA, Nikolov SI, Gammelmark KL, Pedersen MH. Synthetic aperture ultrasound imaging. Ultrasonics. 2006;44:5–15

  25. [33]

    Synthetic tracked aperture ultrasound imaging: design, simulation, and experimental evaluation

    Zhang HK, Cheng A, Bottenus N, Guo X, Trahey GE, Boctor EM. Synthetic tracked aperture ultrasound imaging: design, simulation, and experimental evaluation. J. Med. Imaging. 2016;3:027001. Page 25 of 34

  26. [34]

    Synthetic aperture ladar imaging demonstrations and information at very low return levels

    Barber ZW, Dahl JR. Synthetic aperture ladar imaging demonstrations and information at very low return levels. Appl. Opt. 2014;53:5531–5537

  27. [35]

    Synthetic aperture lidar as a future tool for earth observation

    Turbide S, Marchese L, Terroux M, Bergeron A. Synthetic aperture lidar as a future tool for earth observation. Proc. SPIE. 2017;10563:1115-1122

  28. [36]

    Kinect based real-time synthetic aperture imaging through occlusion

    Yang T, Ma W, Wang S, Li J, Yu J, Zhang Y. Kinect based real-time synthetic aperture imaging through occlusion. Multimed. Tools Appl. 2016;75:6925–6943

  29. [37]

    Occluded-object 3D reconstruction using camera array synthetic aperture imaging

    Pei Z, Li Y, Ma M, Li J, Leng C, Zhang X, Zhang Y. Occluded-object 3D reconstruction using camera array synthetic aperture imaging. Sensors. 2019;19:607

  30. [38]

    Synthetic aperture radio telescopes

    Levanda R, Leshem A. Synthetic aperture radio telescopes. IEEE Signal Process. Mag. 2010;27:14–29

  31. [39]

    Optical aperture synthesis with electronically connected telescopes

    Dravins D, Lagadec T, Nuñez PD. Optical aperture synthesis with electronically connected telescopes. Nat. Commun. 2015;6:6852

  32. [40]

    Airborne Optical Sectioning

    Kurmi I, Schedl DC, Bimber O. Airborne Optical Sectioning. J. Imaging. 2018;4:102

  33. [41]

    Synthetic aperture imaging with drones

    Bimber O, Kurmi I, Schedl DC. Synthetic aperture imaging with drones. IEEE Computer Graphics and Applications. 2019;39:8 – 15

  34. [42]

    A statistical view on synthetic aperture imaging for occlusion removal

    Kurmi I, Schedl DC, Bimber O. A statistical view on synthetic aperture imaging for occlusion removal. IEEE Sensors J. 2019;19:9374 – 9383

  35. [43]

    Thermal airborne optical sectioning

    Kurmi I, Schedl DC, Bimber O. Thermal airborne optical sectioning. Remote Sens. 2019;11:1668

  36. [44]

    Airborne Optical Sectioning for Nesting Observation

    Schedl DC, Kurmi I, Bimber O. Airborne Optical Sectioning for Nesting Observation. Sci Rep. 2020;10:7254

  37. [45]

    Fast automatic visibility optimization for thermal synthetic aperture visualization

    Kurmi I, Schedl DC, Bimber O. Fast automatic visibility optimization for thermal synthetic aperture visualization. IEEE Geosci. Remote Sens. 2020;18:836-840

  38. [46]

    Pose Error Reduction for Focus Enhancement in Thermal Synthetic Aperture Visualization

    Kurmi I, Schedl DC, Bimber O. Pose Error Reduction for Focus Enhancement in Thermal Synthetic Aperture Visualization. IEEE Geosci. Remote Sens. 2021;19:1-5

  39. [47]

    Search and rescue with airborne optical sectioning

    Schedl DC, Kurmi I, Bimber O. Search and rescue with airborne optical sectioning. Nat. Mach. Intell. 2020;2:783-790

  40. [48]

    An autonomous drone for search and rescue in forests using airborne optical sectioning

    Schedl DC, Kurmi I, Bimber O. An autonomous drone for search and rescue in forests using airborne optical sectioning. Sci. Robot. 2021;6(55):1188

  41. [49]

    Combined Person Classification with Airborne Optical Sectioning

    Kurmi I, Schedl DC, Bimber O. Combined Person Classification with Airborne Optical Sectioning. Nature Scientific Reports. 2022;12:3804

  42. [50]

    Through-Foliage Tracking with Airborne Optical Sectioning

    Nathan RJAA, Kurmi I, Schedl DC, Bimber O. Through-Foliage Tracking with Airborne Optical Sectioning. J. Remote Sens. 2022;2022:9812765

  43. [51]

    Inverse Airborne Optical Sectioning

    Nathan RJAA, Kurmi I, Bimber O. Inverse Airborne Optical Sectioning. Drones. 2022;6:9. Page 26 of 34

  44. [52]

    Evaluation of Color Anomaly Detection in Multispectral Images For Synthetic Aperture Sensing, Eng

    Seits F, Kurmi I, Bimber O. Evaluation of Color Anomaly Detection in Multispectral Images For Synthetic Aperture Sensing, Eng. 2022;3(4):541-553

  45. [53]

    Drone swarm strategy for the detection and tracking of occluded targets in complex environments

    Nathan RJAA, Kurmi I, Bimber O. Drone swarm strategy for the detection and tracking of occluded targets in complex environments. Commun Eng. 2023;2:55

  46. [54]

    Synthetic Aperture Anomaly Imaging for Through-Foliage Target Detection

    Nathan RJAA, Bimber O. Synthetic Aperture Anomaly Imaging for Through-Foliage Target Detection. Remote Sens. 2023;15:18

  47. [55]

    Stereoscopic Depth Perception Through Foliage

    Kerschner R, Nathan RJAA, Mantiuk R, Bimber O. Stereoscopic Depth Perception Through Foliage. Nature Scientific Reports. 2024;14:23056

  48. [56]

    Fusion of Single and Integral Multispectral Aerial Images

    Youssef M, Bimber O. Fusion of Single and Integral Multispectral Aerial Images. Remote Sens. 2024;16(4):673

  49. [57]

    Reciprocal Visibility for Guided Occlusion Removal With Drones

    Nathan RJAA, Strand S, Shutin D, Bimber O. Reciprocal Visibility for Guided Occlusion Removal With Drones. IEEE Geosci. Remote Sens. 2024;21:1–5

  50. [58]

    DeconvolutionLab2: An open-source software for deconvolution microscopy

    Sage D, Lauréne D, Ferréol S, Denis F, Guillaume S, Arne S, Romain G, Cédric V, Michael U. DeconvolutionLab2: An open-source software for deconvolution microscopy. Methods. 2017;115:28-41

  51. [59]

    Three-dimensional imaging by deconvolution microscopy

    McNally JG, Karpova T, Cooper J, Conchello JA. Three-dimensional imaging by deconvolution microscopy. Methods. 1999;19(3):373–385

  52. [60]

    Deconvolution methods for 3-D fluorescence microscopy images, IEEE Signal Processing Mag

    Sarder P, Nehorai A. Deconvolution methods for 3-D fluorescence microscopy images, IEEE Signal Processing Mag. 2006;23(3):32–45

  53. [61]

    Blind depth-variant deconvolution of 3D data in wide-field fluorescence microscopy, Sci

    Kim B, Naemura T. Blind depth-variant deconvolution of 3D data in wide-field fluorescence microscopy, Sci. Rep. 2015;5:9894

  54. [62]

    A novel variational approach for multiphoton microscopy image restoration: from PSF estimation to 3D deconvolution

    Ajdenbaum J, Chouzenoux E, Lefort C, Martin S, Pesquet JC. A novel variational approach for multiphoton microscopy image restoration: from PSF estimation to 3D deconvolution. Inverse Probl. 2024;40(6):065003

  55. [63]

    Depth and DOF Cues Make A Better Defocus Blur Detector

    Jin Y, Qian M, Xiong J, Xue N, Xia GS. Depth and DOF Cues Make A Better Defocus Blur Detector. Paper presented at: ICME 2023. Proceeding of IEEE International Conference on Multimedia and Expo; 2023 Jul 10; Brisbane, Australia

  56. [64]

    Depth from Defocus vs

    Schechner YY, Kiryati N. Depth from Defocus vs. Stereo: How Different Really Are They?. International Journal of Computer Vision. 2000;39:141–162

  57. [65]

    A state-of-the-art review of image motion deblurring techniques in precision agriculture

    Huihui Y, Daoliang L, Yingyi C. A state-of-the-art review of image motion deblurring techniques in precision agriculture. Heliyon. 2023;9:6

  58. [66]

    Correction of out-of-focus microscopic images by deep learning

    Zhang C, Jiang H, Liu W, Li J, Tang S, Juhas M, Zhang Y. Correction of out-of-focus microscopic images by deep learning. Computational and Structural Biotechnology Journal. 2022;20:1957–1966

  59. [67]

    Monitoring vegetation systems in the Great Plains with ERTS

    Rouse JW, Haas RH, Schell JA, Deering DW. Monitoring vegetation systems in the Great Plains with ERTS. NASA Spec. Publ. 1974;351(1):309. Page 27 of 34

  60. [68]

    Monitoring forest structure to guide adaptive management of forest restoration: a review of remote sensing approaches

    Camarretta N, Harrison PA, Bailey T, Potts B, Lucieer A, Davidson N, Hunt M. Monitoring forest structure to guide adaptive management of forest restoration: a review of remote sensing approaches. New Forests. 2020;51(4):573-596

  61. [69]

    Global patterns and climatic controls of forest structural complexity

    Ehbrecht M, Seidel D, Annighöfer P, Kreft H, Köhler M, Zemp DC, Puettmann K, Nilus R, Babweteera F, Willim K, Stiers M. Global patterns and climatic controls of forest structural complexity. Nat. Commun. 2021;12(1):519

  62. [70]

    Mahecha MD, Bastos A, Bohn FJ, Eisenhauer N, Feilhauer H, Hartmann H, Hickler T, Kalesse-Los H, Migliavacca M, Otto FE, Peng J, Biodiversity loss and climate extremes— study the feedbacks. Nature. 2022;612(7938):30-32

  63. [71]

    Improved allometric models to estimate the aboveground biomass of tropical trees

    Chave J, Réjou‐Méchain M, Búrquez A, Chidumayo E, Colgan MS, Delitti WBC, Duque A, Eid T, Fearnside PM, Goodman RC, Henry M. Improved allometric models to estimate the aboveground biomass of tropical trees. Global change biology, 2014; 20(10):3177- 3190

  64. [72]

    Mapping carbon accumulation potential from global natural forest regrowth

    Cook-Patton SC, Leavitt SM, Gibbs D, Harris NL, Lister K, Anderson-Teixeira KJ, Briggs RD, Chazdon RL, Crowther TW, Ellis PW, Griscom HP. Mapping carbon accumulation potential from global natural forest regrowth. Nature. 2020; 585:545–550

  65. [73]

    Vegetation stands biomass and carbon stock estimation using NDVI-Landsat 8 imagery in mixed garden of Rancakalong

    Malik AD, Nasrudin A, Withaningsih S. Vegetation stands biomass and carbon stock estimation using NDVI-Landsat 8 imagery in mixed garden of Rancakalong. Paper presented at: IOP Conference Series: Earth and Environmental Science 2023. Proceeding of IOP Publishing; 2023 jul; Sum...

  66. [74]

    Biodiversity loss and climate extremes—study the feedbacks

    Mahecha MD, Bastos A, Bohn FJ, Eisenhauer N, Feilhauer H, Hartmann H, Hickler T, Kalesse-Los H, Migliavacca M, Otto FEL, Peng J. Biodiversity loss and climate extremes—study the feedbacks. Nature. 2022;612(7938):30-32

  67. [75]

    Emerging signals of declining forest resilience under climate change

    Forzieri G, Dakos V, McDowell NG, Ramdane A, Cescatti A. Emerging signals of declining forest resilience under climate change. Nature. 2022; 608(7923):534-539

  68. [76]

    BeyondPixels: A comprehensive review of the evolution of neural radiance fields

    Rabby AKM, Zhang C. BeyondPixels: A comprehensive review of the evolution of neural radiance fields. arXiv. 2023. https://arxiv.org/pdf/2306.03000

  69. [77]

    A survey on 3d gaussian splatting

    Chen G, Wang W. A survey on 3d gaussian splatting. arXiv. 2024. https://arxiv.org/pdf/2401.03890

  70. [78]

    A new method for voxel‐based modelling of three‐dimensional forest scenes with integration of terrestrial and airborne LiDAR data

    Li W, Hu X, Su Y, Tao S, Ma Q, Guo Q. A new method for voxel‐based modelling of three‐dimensional forest scenes with integration of terrestrial and airborne LiDAR data. Methods in Ecology and Evolution. 2024;15(3):569-582

  71. [79]

    Sensitivity of voxel-based estimations of leaf area density with terrestrial LiDAR to vegetation structure and sampling limitations: A simulation experiment

    Soma M, Pimont F, Dupuy JL. Sensitivity of voxel-based estimations of leaf area density with terrestrial LiDAR to vegetation structure and sampling limitations: A simulation experiment. Remote Sens. Environ. 2021;257:112354

  72. [80]

    Beyond Vegetation: A Review Unveiling Additional Insights into Agriculture and Forestry through the Application of Vegetation Indices

    Vélez S, Martínez-Peña R, Castrillo D. Beyond Vegetation: A Review Unveiling Additional Insights into Agriculture and Forestry through the Application of Vegetation Indices. J-Multidisciplinary Scientific Journal. 2023;6(3):421-436. Page 28 of 34

  73. [81]

    Multispectral Light Detection and Ranging Technology and Applications: A Review

    Takhtkeshha N, Mandlburger G, Remondino F, Hyyppä J. Multispectral Light Detection and Ranging Technology and Applications: A Review. Sensors. 2024;24:1669

  74. [82]

    Combining multispectral imagery and synthetic aperture radar for detecting deforestation

    Reinisch EC, Ziemann A, Flynn EB, Theiler J. Combining multispectral imagery and synthetic aperture radar for detecting deforestation. In Algorithms, Technologies, and Applications for Multispectral and Hyperspectral Imagery XXVI. 2020;11392:72-85

  75. [83]

    Novel Algorithms for Remote Estimation of Vegetation Fraction

    Gitelson AA, Kaufman YJ, Stark R, Rundquist D. Novel Algorithms for Remote Estimation of Vegetation Fraction. Remote Sens. Environ. 2002;80:76–87

  76. [84]

    Under-Canopy Drone 3D Surveys for Wild Fruit Hotspot Mapping

    Trybała P, Morelli L, Remondino F, Farrand L, Couceiro MS. Under-Canopy Drone 3D Surveys for Wild Fruit Hotspot Mapping. Drones. 2024;8(10):577

  77. [85]

    Forest in situ observations through a fully automated under-canopy unmanned aerial vehicle

    Liang X, Yao H, Qi H, Wang X. Forest in situ observations through a fully automated under-canopy unmanned aerial vehicle. Geo-Spatial Information Science. 2024;27(4):983–999

  78. [86]

    A Drone-based Prototype Design and Testing for Under- the-canopy Imaging and Onboard Data Analytics

    Zanone RO, Liu T, Velni JM. A Drone-based Prototype Design and Testing for Under- the-canopy Imaging and Onboard Data Analytics. IFAC-PapersOnLine. 2022;55(32):171- 176

  79. [87]

    Exploring the potential of transmittance vegetation indices for leaf functional traits retrieval

    Chen Y, Sun J, Wang L, Shi S, Qiu F, Gong W, Tagesson T. Exploring the potential of transmittance vegetation indices for leaf functional traits retrieval. GIScience & Remote Sensing. 2023;60(1):2168410

  80. [88]

    Review of indirect optical measurements of leaf area index: Recent advances, challenges, and perspectives

    Yan G, Hu R, Luo J, Weiss M, Jiang H, Mu X, Xie D, Zhang W. Review of indirect optical measurements of leaf area index: Recent advances, challenges, and perspectives. Agricultural and Forest Meteorology. 2019;265:390-411. Page 29 of 34 Supplementary Materials for DeepForest: S...

  81. [2016]

    Proceeding of IEEE Conference on Computer Vision and Pattern Recognition;2016 Dec 12; Las Vegas, NV, USA

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.