Pith. sign in

REVIEW 4 major objections 6 minor 67 references

Towards Interpretable PolSAR Image Classification: Polarimetric Scattering Mechanism Informed Concept Bottleneck and Kolmogorov-Arnold Network

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read The paper claims PolSAR deep classifiers can be opened up by routing features through physical scattering concepts, then mapping concepts to labels with a Kolmogorov-Arnold network that yields closed-form formulas, without losing…

desk verdict A useful first CBM/KAN application to PolSAR whose main interpretability claim is undercut by an unspecified concept-labeling procedure and missing fidelity checks. read the letter →

arxiv 2507.03315 v1 pith:3SGKMVBB submitted 2025-07-04 eess.IV cs.CV

Jinqi Zhang , Fangzhou Han , Di Zhuang , Lamei Zhang , Bin Zou , Li Yuan This is my paper
classification eess.IVcs.CV
keywords PolSARimageclassificationconceptbottleneckmodelpolarimetrictargetdecompositionscatteringmechanismKolmogorov-Arnoldnetworkinterpretabledeeplearningsymbolicformularemotesensing
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 tries to open the black box of deep-learning PolSAR image classification. It replaces the usual final feature-to-label mapping with two transparent stages: a concept bottleneck layer that predicts human-comprehensible labels such as 'double-bounce scattering dominant' or 'high polarization entropy,' and a Kolmogorov-Arnold network that combines those concept predictions into a class decision. Because the concept labels come from polarimetric target decomposition, the intermediate layer is tied to physically verifiable scattering mechanisms rather than arbitrary semantics. The authors show on three datasets that this design reaches accuracy comparable to or better than conventional deep models, and they report that the concept-to-label mapping can be written as closed-form symbolic formulas. If this holds, the model's reasoning is traceable end to end: raw polarimetric data to physical concepts to an explicit decision rule.

What carries the argument

The load-bearing object is the polarimetric concept label set, built from Cloude–Pottier ($H$, $\alpha$, $A$), Freeman–Durden (surface, double-bounce, volume components), and Huynen (nine target parameters) decompositions. These decompositions turn the $3\times 3$ coherency matrix into physically meaningful attributes, and the paper discretizes those attributes into binary concepts such as 'surface scattering dominant' and 'high polarization entropy.' The PaCBM structure runs the concept branch in parallel with the main classifier over a shared CNN and vision transformer encoder with cross-attention, so the concept bottleneck supervises feature learning without forcing all information through it. The KAN then maps predicted concepts to class logits via learned univariate spline functions, which is what enables the symbolic readout.

What would settle it

Take any test pixel, compute its PTD parameters directly from its coherency matrix, and compare them with the concept scores the network predicts for that pixel; if pixels with high predicted 'double-bounce scattering dominant' do not actually have dominant Freeman–Durden double-bounce power, or if systematically mislabeled concept training patches are demonstrably the source of concept errors, then the bottleneck is not actually encoding the claimed physics.

Watch

Extended reading notes

Core claim

The central claim is that interpretability does not have to be bought with accuracy: a Parallel Concept Bottleneck Network built on PTD-derived concept labels plus a Kolmogorov-Arnold network can classify PolSAR images while exposing the decision process. The parallel branch keeps a standard feature-to-label path for accuracy while a concept branch learns to predict PTD-based concepts, with a shared encoder and a weighted concept loss. Replacing the MLP with KAN lets the concept-to-label map be expressed as sums of B-spline univariate functions; the paper converts selected KAN nodes into human-readable formulas, such as the logit for 'Mountain' expressed as combinations of sine, exponential, polynomial, and linear terms in concept variables. The authors report average accuracies of 92.16% on GF3-SF, 97.28% on RS2-SF, and 95.51% on Oberpfaffenhofen, and demonstrate intervention: flipping a wrongly activated concept from 0.84 to 0 changes a misclassified sample's label to the correct 'Developed' class.

Load-bearing premise

The paper's interpretability evidence rests on the assumption that the concept labels it hand-built from six 200x200 patches of one GF-3 image are correct, complete, and valid for all three datasets; if those labels do not reflect real scattering behavior or omit concepts the network actually uses, the concept predictions and symbolic formulas explain the labels, not the model's true reasoning.

Editorial extensions

If this is right

  • A user can audit any single classification by reading the predicted concept scores and checking them against physical expectations for that terrain type.
  • Because the concept-to-label stage is written as symbolic formulas, decision rules can be inspected, simplified, and compared across sensors without re-running the network.
  • Concept intervention becomes a practical debugging tool: manually correcting an erroneous concept activation should change the output to the intended class, as shown for a 'Developed' sample.
  • The parallel branch means concept supervision can regularize feature learning instead of forcing all information through the bottleneck, so interpretability does not necessarily cost accuracy.
  • If the concept vocabulary is band-robust as claimed, the same PTD-derived concepts can supervise classification at both C-band and L-band without rebuilding labels.

Reading between the lines

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

  • The same recipe could be applied to other physical priors: using interferometric phase or coherence descriptors as concepts would extend this style of auditability to InSAR and PolInSAR models.
  • A direct extension is to report per-concept prediction accuracy; if certain concepts are hard to predict, the symbolic formulas may be assigning weight to concept values the network does not actually estimate reliably.
  • Cross-sensor transfer audits are a natural next test: applying the GF-3-derived concept vocabulary to L-band data would reveal whether concept disagreements encode genuine frequency-dependent scattering or just label noise.
  • One could also test the faithfulness of the extracted formulas by removing individual concepts and comparing the formula's sensitivity with the network's actual response, which the paper does not report.
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

4 major / 6 minor

Summary. The manuscript proposes PaCBM-KAN, a PolSAR image classification framework that couples a parallel concept bottleneck network with Kolmogorov-Arnold networks. Concepts are defined from polarimetric target decomposition (Cloude-Pottier, Freeman-Durden, Huynen) and are claimed to provide physically meaningful intermediate representations. The model is evaluated on GF3-SF, RS2-SF, and Oberpfaffenhofen, reporting competitive accuracy (AA 92.16%, 97.28%, and 95.51%, respectively). Interpretability is demonstrated through concept prediction visualization, KAN-based symbolic formulas for concept-to-label mapping, and concept intervention for correcting misclassifications. The central claim is that high-dimensional PolSAR features can be transformed into human-comprehensible, physically grounded concepts while preserving classification accuracy and enabling traceable decision-making.

Significance. If the interpretability claim is fully supported, the paper would be a meaningful contribution to interpretable PolSAR classification: it is, to my knowledge, the first to combine PTD-derived concept bottlenecks with KAN, and it reports accuracy competitive with or above several strong baselines while offering concept prediction and intervention. The paper also includes a useful ablation of CBM versus PaCBM and KAN versus MLP, as well as a hyperparameter study for lambda and KAN structure. However, three evidentiary gaps currently limit the significance: the per-pixel concept-label generation procedure is not specified, the symbolic formulas in Table VI are presented without any fidelity check, and the cross-dataset transfer of concept labels is asserted rather than validated. These gaps are fixable, so the work is promising but not yet established.

major comments (4)
  1. [Section III-B, Table I, Eq. (13)] The per-pixel assignment of concept labels to training samples is not specified. The text describes manually selecting 200x200 patches of six GF-3 classes and summarizes concept labels in Table I, but no algorithm is given that converts the 9-dimensional T-matrix input of each training sample into the concept vector c_n used in D_c. If each sample is simply assigned the Table I prototype of its class, then the concept loss in Eq. (13) is a second category-level classification loss, and the concept predictions in Fig. 8 demonstrate category recognition rather than scattering-mechanism grounding. Please provide the exact per-pixel labeling procedure, or clarify how the Table I prototypes are expanded to every pixel, and validate the resulting labels on held-out patches.
  2. [Table VI, Section IV-C-2] The symbolic formulas in Table VI are presented as evidence of traceable decision-making, but the manuscript reports no fidelity metric showing how well these formulas reproduce the trained KAN outputs. A closed-form expression that deviates from the learned function by a large error cannot support the interpretability claim. Please report the approximation error (e.g., R-squared or mean absolute error) between the formula and the KAN logits, over the concept-value ranges encountered in the test set, and describe how the formulas were extracted.
  3. [Section IV-A, Table I] The concept labels are derived solely from six 200x200 patches of GF-3 (C-band) data, yet the same labels are used for RS2-SF and Oberpfaffenhofen. The Oberpfaffenhofen classes (built-up, wood land, open area) do not appear in Table I, and the thresholds in Eqs. (9)-(10) and the H/alpha/A interval splits were calibrated on GF-3 without L-band validation. The claim that the scattering mechanisms are band-robust needs direct evidence: please provide per-dataset concept-label statistics or a concept-prediction accuracy evaluation on the validation/test splits of RS2-SF and Oberpfaffenhofen.
  4. [Tables II-V] All classification results are reported from a single run, with no error bars or repeated-seed statistics. The differences among the best-performing methods are often small (e.g., Table III OA 98.04% for HyCVNet versus 98.06% for PaCBM-j, and Kappa 97.20% versus 97.23%). The statements that PaCBM 'outperforms' the baselines are therefore not statistically supported. Please report means and standard deviations over at least three independent runs with different random seeds.
minor comments (6)
  1. [Abstract] The sentence 'the features could be conceptualization' is ungrammatical; it should be 'the features could be conceptualized.'
  2. [Section IV-B-1] The text refers to 'HyCBNet' in the GF3-SF and Oberpfaffenhofen analyses, but the tables and references use 'HyCVNet'; please make the naming consistent.
  3. [Section IV-A] The Oberpfaffenhofen dataset name is spelled 'Oberpfaffenhofe' in the dataset description; the standard spelling is 'Oberpfaffenhofen.'
  4. [Section III-B] Equation (6) is typeset in a way that is hard to read: the definitions of log3 and alpha_n are not clearly separated from the formula, and the notation for P_n and alpha_n deserves explicit definitions before the equation.
  5. [References] References [30] and [60] are duplicates (both are Alkhatib's HybridCVNet paper); renumber to cite the work only once.
  6. [Section IV-C-2] The caption of Table VI does not define the concept indices c1...c33 or state that these correspond to the ordering in Table I; please add a mapping so the formulas can be interpreted.

Circularity Check

2 steps flagged · score 6.0 of 10

Concept labels are per-class prototypes, so concept prediction and intervention evidence reduce to category classification; the symbolic formulas are post hoc fits of that same class-prototype mapping.

  1. self definitional [Section III-B/III-C, Eqs. (12)-(13), Table I, Fig. 8]
    "Using 0/1 to represent the correspondence between different categories and concept sets, the original dataset D can be expanded to Dc = {Xn, yn, cn}, where the cn ∈ RK is the vector of the concept labels."

    Table I supplies exactly one concept-label vector per class (Mountain, Water, Vegetation, etc.), and the sentence quoted above defines cn via a 0/1 correspondence between categories and concept sets. Thus every training sample of a given class receives its class's fixed prototype as cn. The concept predictor Kf→c is trained with BCE loss (Eq. 13) to reproduce these prototypes, so its outputs are class-prototype predictions by construction. The interpretability demonstrations in Fig. 8 ('precise conceptual analysis of previously unseen samples') and Fig. 10 (intervention correcting High-Density Urban to Developed) therefore show the model matching its training prototypes, not an independently verified scattering-mechanism grounding.

  2. fitted input called prediction [Section IV-C-2, Table VI]
    "Table VI presents symbolic formulas derived using KAN, representing the conceptual mapping from land cover semantics to classification labels."

    These symbolic formulas are post hoc fits to the trained Kc→t, which was itself supervised by the same category-derived concept vectors from Table I. The paper presents the result as 'the analytical function for predicting category labels from conceptual labels,' but the formula is a fitted re-expression of a classifier whose inputs are per-class prototypes. It inherits the circularity of the concept-label construction: the formulas map class prototypes to class labels, not independently measured PTD concepts to labels, and they provide no out-of-sample validation of the physical interpretability claim.

full rationale

The classification-accuracy results are not circular in the target-label sense: the model is trained with standard class labels and PTD-derived auxiliary supervision, and the reported AA/OA/Kappa numbers are measured on held-out data against independent baselines. However, the paper's own description makes the concept-label vector a per-class constant. Section III-B derives one concept vector per category from six manually selected 200x200 GF-3 patches; Section III-C expands D to Dc with cn as 'the vector of the concept labels' using 0/1 correspondence between categories and concept sets; Table I then lists one vector per class. Consequently, Kf→c and Kc→t are trained to map features to class prototypes and prototypes to classes. The interpretability demonstrations (Fig. 8 concept predictions, Fig. 10 intervention, Table VI formulas) therefore show the model matching its training prototypes, not an independent physical-mechanism verification. This is a partial circularity of the interpretability claim: the derivation chain reduces by construction to category classification at the concept level. No load-bearing self-citations exist: the authors' prior contrastive-learning work [44] and the HyCVNet-inspired backbone [60] are peripheral, and no uniqueness theorem is imported. Score 6 reflects that the central interpretability demonstration reduces by construction, while the accuracy numbers and the PTD physics themselves retain independent content.

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

No new physical particles, forces, or dimensions are introduced. The only new objects are architectural components such as the PaCBM parallel branch and the KAN-based mapping, which are trained artifacts rather than postulated entities.

free parameters (4)
  • Regularization weight lambda = 0.7
    Controls the balance between main and concept losses in Eq. (20). Selected by a grid search on GF3-SF and then fixed for all datasets, so conclusions about interpretability and accuracy depend on this choice.
  • KAN grid size and spline order = grid size 7, spline order 3
    Selected using average accuracy and parameter count on the GF3-SF dataset. This choice shapes the symbolic formulas and the claimed interpretability of the concept-to-label mapping.
  • KAN hidden layer width = 16 nodes
    A two-layer KAN with 16 hidden nodes is used for concept-to-label mapping. This is an architecture choice that affects the complexity and form of the extracted formulas.
  • Concept label thresholds = Pmax >= 0.4, Pmax - P2nd >= 0.1, Pi >= 0.5 Pmax
    Equations (9) and (10) define dominant, secondary, and weak scattering concepts. These hand-chosen thresholds determine all Freeman-Durden concept labels and therefore shape the concept supervision signal.
assumptions (4)
  • domain assumption Polarimetric target decompositions (Cloude-Pottier, Freeman-Durden, Huynen) produce physically meaningful scattering mechanism descriptions.
    Section III-A assumes these established decompositions correctly characterize terrain scattering, which is the foundation for constructing concept labels.
  • domain assumption Concept labels constructed from six classes on GF-3 are correct and general enough to serve as supervision for RS2-SF and Oberpfaffenhofen.
    Section III-B builds concept labels from manually selected patches on one sensor and then applies the same labels to other sensors and frequency bands without independent validation.
  • standard math The Kolmogorov-Arnold Representation Theorem justifies replacing MLP layers with a sum of univariate spline functions.
    Section III-C invokes KART as the theoretical basis for KAN, which is a standard mathematical theorem.
  • domain assumption The dual-branch CNN-ViT encoder inspired by HyCVNet is a sufficient feature extractor for the proposed framework.
    Section III-D adopts the hybrid encoder without an ablation study of the encoder itself, so the performance and interpretability results depend on this architecture choice.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Towards Interpretable PolSAR Image Classification: Polarimetric Scattering Mechanism Informed Concept Bottleneck and Kolmogorov-Arnold Network." pith.science (2026). https://pith.science/paper/3SGKMVBB

@misc{pith2026250703315,
  author       = {Pith},
  title        = {Pith review of: Towards Interpretable PolSAR Image Classification: Polarimetric Scattering Mechanism Informed Concept Bottleneck and Kolmogorov-Arnold Network},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3SGKMVBB}},
  note         = {Machine review of arXiv:2507.03315}
}
read the original abstract

In recent years, Deep Learning (DL) based methods have received extensive and sufficient attention in the field of PolSAR image classification, which show excellent performance. However, due to the ``black-box" nature of DL methods, the interpretation of the high-dimensional features extracted and the backtracking of the decision-making process based on the features are still unresolved problems. In this study, we first highlight this issue and attempt to achieve the interpretability analysis of DL-based PolSAR image classification technology with the help of Polarimetric Target Decomposition (PTD), a feature extraction method related to the scattering mechanism unique to the PolSAR image processing field. In our work, by constructing the polarimetric conceptual labels and a novel structure named Parallel Concept Bottleneck Networks (PaCBM), the uninterpretable high-dimensional features are transformed into human-comprehensible concepts based on physically verifiable polarimetric scattering mechanisms. Then, the Kolmogorov-Arnold Network (KAN) is used to replace Multi-Layer Perceptron (MLP) for achieving a more concise and understandable mapping process between layers and further enhanced non-linear modeling ability. The experimental results on several PolSAR datasets show that the features could be conceptualization under the premise of achieving satisfactory accuracy through the proposed pipeline, and the analytical function for predicting category labels from conceptual labels can be obtained by combining spline functions, thus promoting the research on the interpretability of the DL-based PolSAR image classification model.

Figures

Figures reproduced from arXiv: 2507.03315 by the authors.

Figure 1
Figure 1. The conceptual labels constructed by the PTD of six typical terrain surface categories in GF-3 dataset. (a) Pixel block selection for PTD (b) Cloude [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. The overall model of the proposed method. (a) the proposed PaCBM. The red data stream represents the calculation of category labels directly [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Classification results of the whole map on GF3 San Francisco dataset. (a) GF3 label. (b) Result of SF-CNN. (c) Result of ViT. (d) Result of PaCBM. [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Classification results of the whole map on RS2 San Francisco dataset. (a) RS2 label. (b) Result of SF-CNN. (c) Result of ViT. (d) Result of PaCBM. [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]
Figure 5
Figure 5. Figure 5: Classification results of the whole map on Oberpfaffenhofen dataset. (a) Oberpfaffenhofen label. (b) Result of SF-CNN. (c) Result of ViT. (d) Result [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]
Figure 6
Figure 6. Figure 6: Effect of hyperparameter λ on AA in the GF3-SF dataset results, the hyperparameter λ significantly influences the per￾formance of PaCBM-joint. The observed trend shows that the AA initially increases with λ reaching a peak before declining as λ continues to grow. Relat…
Figure 7
Figure 7. Figure 7: Impact of grid size and spline order on AA and params in KAN [PITH_FULL_IMAGE:figures/full_fig_p012_7.png]
Figure 8
Figure 8. Figure 8: Some results of the concept prediction in PolSAR dataset. We selected [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]
Figure 9
Figure 9. Figure 9: Visualization of Internal Computation and Comparison between KAN and MLP. (a) MLP performs mapping via fixed activation functions combined [PITH_FULL_IMAGE:figures/full_fig_p013_9.png]
Figure 10
Figure 10. Figure 10: Example of explainable intervention and classification correction [PITH_FULL_IMAGE:figures/full_fig_p014_10.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

67 extracted references · 61 canonical work pages

  1. [1]

    Tsunami damage investigation of built-up areas using multitemporal spaceborne full polarimetric sar images,

    S.-W. Chen and M. Sato, “Tsunami damage investigation of built-up areas using multitemporal spaceborne full polarimetric sar images,” IEEE Trans. Geosci. Remote Sens., vol. 51, no. 4, pp. 1985–1997, 2013

    work page 1985

  2. [2]

    Generation of digital terrain models with a ground-based sar system,

    G. Nico, D. Leva, J. Fortuny-Guasch, G. Antonello, and D. Tarchi, “Generation of digital terrain models with a ground-based sar system,” IEEE Trans. Geosci. Remote Sens. , vol. 43, no. 1, pp. 45–49, 2005

    work page 2005

  3. [3]

    Quantitative comparison of classi- fication capability: fully polarimetric versus dual and single-polarization sar,

    J.-S. Lee, M. Grunes, and E. Pottier, “Quantitative comparison of classi- fication capability: fully polarimetric versus dual and single-polarization sar,” IEEE Trans. Geosci. Remote Sens., vol. 39, no. 11, pp. 2343–2351, 2001

    work page 2001

  4. [4]

    Polarimetric convo- lutional network for polsar image classification,

    X. Liu, L. Jiao, X. Tang, Q. Sun, and D. Zhang, “Polarimetric convo- lutional network for polsar image classification,” IEEE Trans. Geosci. Remote Sens., vol. 57, no. 5, pp. 3040–3054, 2018

    work page 2018

  5. [5]

    Classification of polarimetric sar images using compact convolutional neural networks,

    M. Ahishali, S. Kiranyaz, T. Ince, and M. Gabbouj, “Classification of polarimetric sar images using compact convolutional neural networks,” Remote Sens., vol. 58, no. 1, pp. 28–47, 2021

    work page 2021

  6. [6]

    Polsarnet: A deep fully convolutional network for polarimetric sar image classification,

    A. G. Mullissa, C. Persello, and A. Stein, “Polsarnet: A deep fully convolutional network for polarimetric sar image classification,” IEEE J. Sel. Topics Appl. Earth Observ. Remote Sens. , vol. 12, no. 12, pp. 5300–5309, 2019

    work page 2019

  7. [7]

    Learning rotation domain deep mutual information using convolutional lstm for unsupervised polsar image classification,

    L. Wang, X. Xu, R. Gui, R. Yang, and F. Pu, “Learning rotation domain deep mutual information using convolutional lstm for unsupervised polsar image classification,” Remote Sens. , vol. 12, no. 24, p. 4075, 2020

    work page 2020

  8. [8]

    Exploring convolutional lstm for polsar image classification,

    L. Wang, X. Xu, H. Dong, R. Gui, R. Yang, and F. Pu, “Exploring convolutional lstm for polsar image classification,” in Proc. 2018 IEEE Int. Geosci. Remote Sens. Symp. (IGARSS) . IEEE, 2018, pp. 8452– 8455

    work page 2018

Show all 67 references
  1. [9]

    Multi-frequency polsar image classification based on cross attention vit,

    H. Wang, D. Ren, J. Yin, and J. Yang, “Multi-frequency polsar image classification based on cross attention vit,” in Proc. IET Int. Radar Conf. (IRC), vol. 2023. IET, 2023, pp. 3219–3222

    2023

  2. [10]

    Land cover classification for polarimetric sar images based on vision transformer,

    H. Wang, C. Xing, J. Yin, and J. Yang, “Land cover classification for polarimetric sar images based on vision transformer,” Remote Sens. , vol. 14, no. 18, p. 4656, 2022

    2022

  3. [11]

    Remote sensing image interpretation with semantic graph-based meth- ods: A survey,

    S. Sun, S. Dustdar, R. Ranjan, G. Morgan, Y . Dong, and L. Wang, “Remote sensing image interpretation with semantic graph-based meth- ods: A survey,” IEEE J. Sel. Topics Appl. Earth Observ. Remote Sens. , vol. 15, pp. 4544–4558, 2022

    2022

  4. [12]

    Metaseg- net: Metadata-collaborative vision-language representation learning for semantic segmentation of remote sensing images,

    L. Wang, S. Dong, Y . Chen, X. Meng, S. Fang, and S. Fei, “Metaseg- net: Metadata-collaborative vision-language representation learning for semantic segmentation of remote sensing images,” IEEE Trans. Geosci. Remote Sens., vol. 62, pp. 1–11, 2024

    2024

  5. [13]

    Better visual interpretation for remote sensing scene classification,

    X. Huang, Y . Sun, S. Feng, Y . Ye, and X. Li, “Better visual interpretation for remote sensing scene classification,” IEEE Geosci. Remote Sens. Lett., vol. 19, pp. 1–5, 2022

    2022

  6. [14]

    Uav hyperspectral remote sensing image classification: A systematic review,

    Z. Zhang, L. Huang, Q. Wang, L. Jiang, Y . Qi, S. Wang, T. Shen, B.-H. Tang, and Y . Gu, “Uav hyperspectral remote sensing image classification: A systematic review,” IEEE J. Sel. Topics Appl. Earth Observ. Remote Sens. , vol. 18, pp. 3099–3124, 2025

    2025

  7. [15]

    General polarimetric model-based decomposition for coherency matrix,

    S.-W. Chen, X.-S. Wang, S.-P. Xiao, and M. Sato, “General polarimetric model-based decomposition for coherency matrix,” IEEE Trans. Geosci. Remote Sens., vol. 52, no. 3, pp. 1843–1855, 2014

    2014

  8. [16]

    An extension of a complete model-based decomposition of polarimetric sar data,

    F. Zhu, Y . Zhang, and D. Li, “An extension of a complete model-based decomposition of polarimetric sar data,” IEEE Geosci. Remote Sens. Lett., vol. 13, no. 2, pp. 287–291, 2016. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 15

    2016

  9. [17]

    Polsar image analysis of wetlands using a modified four-component scattering power decomposition,

    Y . Yajima, Y . Yamaguchi, R. Sato, H. Yamada, and W.-M. Boerner, “Polsar image analysis of wetlands using a modified four-component scattering power decomposition,” IEEE Trans. Geosci. Remote Sens. , vol. 46, no. 6, pp. 1667–1673, 2008

    2008

  10. [18]

    Modeling and interpretation of scattering mechanisms in polarimetric synthetic aperture radar: Advances and perspectives,

    S.-W. Chen, Y .-Z. Li, X.-S. Wang, S.-P. Xiao, and M. Sato, “Modeling and interpretation of scattering mechanisms in polarimetric synthetic aperture radar: Advances and perspectives,” IEEE Signal Process. Mag., vol. 31, no. 4, pp. 79–89, 2014

    2014

  11. [19]

    A three-component scattering model for polarimetric sar data,

    A. Freeman and S. L. Durden, “A three-component scattering model for polarimetric sar data,” IEEE Trans. Geosci. Remote. Sens. , vol. 36, pp. 963–973, 1998

    1998

  12. [20]

    Sublook2sublook: A self- supervised speckle filtering framework for single sar images,

    J.-W. Deng, M.-D. Li, and S.-W. Chen, “Sublook2sublook: A self- supervised speckle filtering framework for single sar images,” IEEE Trans. Geosci. Remote Sens. , vol. 62, pp. 1–13, 2024

    2024

  13. [21]

    An active deep learning approach for minimally supervised polsar image classification,

    H. Bi, F. Xu, Z. Wei, Y . Xue, and Z. Xu, “An active deep learning approach for minimally supervised polsar image classification,” IEEE Trans. Geosci. Remote Sens. , vol. 57, no. 11, pp. 9378–9395, 2019

    2019

  14. [22]

    Concept bottleneck models,

    P. W. Koh, T. Nguyen, Y . S. Tang, S. Mussmann, E. Pierson, B. Kim, and P. Liang, “Concept bottleneck models,” in Proc. 32nd Int. Conf. Mach. Learn. (ICML) , 2020, pp. 5338–5348

    2020

  15. [23]

    Editable concept bottleneck models,

    L. Hu, C. Ren, Z. Hu, H. Lin, C.-L. Wang, H. Xiong, J. Zhang, and D. Wang, “Editable concept bottleneck models,” arXiv:2405.15476, 2024

    2024 arXiv

  16. [24]

    Post-hoc concept bottleneck models,

    M. Yuksekgonul, M. Wang, and J. Zou, “Post-hoc concept bottleneck models,” arXiv:2205.15480, 2022

    2022 arXiv

  17. [25]

    Kan: Kolmogorov-arnold networks,

    Z. Liu, Y . Wang, S. Vaidya, F. Ruehle, J. Halverson, M. Solja ˇci´c, T. Y . Hou, and M. Tegmark, “Kan: Kolmogorov-arnold networks,” arXiv:2404.19756, 2024

    2024 arXiv

  18. [26]

    Classification of polsar images by stacked random forests,

    R. H ¨ansch and O. Hellwich, “Classification of polsar images by stacked random forests,” ISPRS Int. J. Geoinf , vol. 7, no. 2, 2018

    2018

  19. [27]

    Polarimetric sar image classification using deep convolutional neural networks,

    Y . Zhou, H. Wang, F. Xu, and Y .-Q. Jin, “Polarimetric sar image classification using deep convolutional neural networks,” IEEE Geosci. Remote Sens. Lett. , vol. 13, no. 12, pp. 1935–1939, 2016

    1935

  20. [28]

    A novel polarimetric sar classi- fication method integrating pixel-based and patch-based classification,

    R. Yang, Z. Hu, Y . Liu, and Z. Xu, “A novel polarimetric sar classi- fication method integrating pixel-based and patch-based classification,” IEEE Geosci. Remote Sens. Lett. , vol. 17, no. 3, pp. 431–435, 2020

    2020

  21. [29]

    Exploring vision transformers for polarimetric sar image classification,

    H. Dong, L. Zhang, and B. Zou, “Exploring vision transformers for polarimetric sar image classification,” IEEE Trans. Geosci. Remote Sens., vol. 60, pp. 1–15, 2022

    2022

  22. [31]

    Polarimetry-inspired contrastive learning for class-imbalanced polsar image classification,

    Z. Kuang, H. Bi, F. Li, C. Xu, and J. Sun, “Polarimetry-inspired contrastive learning for class-imbalanced polsar image classification,” IEEE Trans. Geosci. Remote Sens. , vol. 62, pp. 1–19, 2024

    2024

  23. [32]

    Mcdiff: A multilevel conditional diffusion model for polsar image classification,

    Q. Zhang, X. Fang, T. Liu, R. Wu, L. Liu, and C. He, “Mcdiff: A multilevel conditional diffusion model for polsar image classification,” IEEE J. Sel. Topics Appl. Earth Observ. Remote Sens., vol. 18, pp. 6721– 6737, 2025

    2025

  24. [33]

    Complex-valued 2d-3d hybrid convolutional neural network with attention mechanism for polsar image classification,

    W. Li, H. Xia, J. Zhang, Y . Wang, Y . Jia, and Y . He, “Complex-valued 2d-3d hybrid convolutional neural network with attention mechanism for polsar image classification,” Remote Sens., vol. 16, no. 16, 2024

    2024

  25. [34]

    A lightweight riemannian covariance matrix convolutional network for polsar image classification,

    J. Shi, W. Wang, H. Jin, M. Nie, and S. Ji, “A lightweight riemannian covariance matrix convolutional network for polsar image classification,” IEEE Trans. Geosci. Remote Sens. , 2024

    2024

  26. [35]

    Polsar image classification via a multi-granularity hybrid cnn-vit model with external tokens and cross-attention,

    W. Wang, J. Wang, D. Quan, M. Yang, J. Sun, and B. Lu, “Polsar image classification via a multi-granularity hybrid cnn-vit model with external tokens and cross-attention,” IEEE J. Sel. Topics Appl. Earth Observ. Remote Sens., 2024

    2024

  27. [36]

    Dual-branch cnn incorporating multiscale svd profile for polsar image classification,

    N. Das, A. Bortiew, S. Patra, and L. Bruzzone, “Dual-branch cnn incorporating multiscale svd profile for polsar image classification,” IEEE Trans. Geosci. Remote Sens. , 2024

    2024

  28. [37]

    Krogager, D

    E. Krogager, D. Danmarks Tekniske Hojskole (Lingby, and D. D. R. Establishment, Aspects of Polarimetric Radar Imaging. Danish Defence Research Establishment, 1993

    1993

  29. [38]

    Feature motivated polarization scattering matrix decomposition,

    W. Cameron and L. Leung, “Feature motivated polarization scattering matrix decomposition,” in IEEE Int. Conf. Radar , 1990, pp. 549–557

    1990

  30. [39]

    Four- component scattering model for polarimetric sar image decomposition,

    Y . Yamaguchi, T. Moriyama, M. Ishido, and H. Yamada, “Four- component scattering model for polarimetric sar image decomposition,” IEEE Trans. Geosci. Remote Sens. , vol. 43, pp. 1699–1706, 2005

    2005

  31. [40]

    Combination of support vector machine and h-alpha decomposition for subsurface target classification of gpr,

    H. Zhou, X. Feng, Y . Zhang, E. Nilot, M. Zhang, Z. Dong, and J. Qi, “Combination of support vector machine and h-alpha decomposition for subsurface target classification of gpr,” in Proc. 17th Int. Conf. Ground Penetr. Radar (GPR), 2018, pp. 1–4

    2018

  32. [41]

    An iterative polsar image classification method with utilizing scattering and contextual information,

    M. Imani, “An iterative polsar image classification method with utilizing scattering and contextual information,” Multimed. Tools Appl. , vol. 83, no. 6, pp. 16 605–16 621, 2024

    2024

  33. [42]

    A comparative study on classification features between high-resolution and polarimetric sar images through unsupervised classification methods,

    J. Qu, X. Qiu, W. Wang, Z. Wang, B. Lei, and C. Ding, “A comparative study on classification features between high-resolution and polarimetric sar images through unsupervised classification methods,” Remote Sens., vol. 14, no. 6, 2022

    2022

  34. [43]

    A feature fusion network for polsar image classification based on physical features and deep features,

    W. Hua, Q. Hou, X. Jin, L. Liu, N. Sun, and Z. Meng, “A feature fusion network for polsar image classification based on physical features and deep features,” IEEE Geosci. Remote Sens. Lett. , vol. 21, pp. 1–5, 2024

    2024

  35. [44]

    Mlr-simsiam: A contrastive pretraining model based on polarimetric jittering and mu- tual learning regularizer for polsar image classification,

    L. Zhang, F. Han, T. Li, H. Dai, B. Zou, and H. Dong, “Mlr-simsiam: A contrastive pretraining model based on polarimetric jittering and mu- tual learning regularizer for polsar image classification,” IEEE Geosci. Remote Sens. Lett. , vol. 21, pp. 1–5, 2024

    2024

  36. [45]

    Learning deep features for discriminative localization,

    B. Zhou, A. Khosla, A. Lapedriza, A. Oliva, and A. Torralba, “Learning deep features for discriminative localization,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR) , 2016, pp. 2921–2929

    2016

  37. [46]

    Grad-cam: Visual explanations from deep networks via gradient-based localization,

    R. R. Selvaraju, M. Cogswell, A. Das, R. Vedantam, D. Parikh, and D. Batra, “Grad-cam: Visual explanations from deep networks via gradient-based localization,” in Proc. IEEE Int. Conf. Comput. Vis.(ICCV), 2017, pp. 618–626

    2017

  38. [47]

    Concept embedding models: Beyond the accuracy-explainability trade- off,

    M. Espinosa Zarlenga, P. Barbiero, G. Ciravegna, G. Marra, F. Giannini, M. Diligenti, Z. Shams, F. Precioso, S. Melacci, A. Weller et al. , “Concept embedding models: Beyond the accuracy-explainability trade- off,” Adv. Neural Inf. Process. Syst. , vol. 35, pp. 21 400–21 413, 2022

    2022

  39. [48]

    Label-free concept bottleneck models,

    T. Oikarinen, S. Das, L. M. Nguyen, and T.-W. Weng, “Label-free concept bottleneck models,” arXiv:2304.06129, 2023

    2023 arXiv

  40. [49]

    Towards concept-based interpretability of skin lesion diagnosis using vision-language models,

    C. Patr ´ıcio, L. F. Teixeira, and J. C. Neves, “Towards concept-based interpretability of skin lesion diagnosis using vision-language models,” in Proc. IEEE Int. Symp. Biomed. Imaging (ISBI) . IEEE, 2024, pp. 1–5

    2024

  41. [50]

    Explainable neural network-based modulation classification via concept bottleneck models,

    L. J. Wong and S. McPherson, “Explainable neural network-based modulation classification via concept bottleneck models,” in Proc. IEEE Annu. Comput. Commun. Workshop Conf. (CCWC) . IEEE, 2021, pp. 0191–0196

    2021

  42. [51]

    Explain- able, physics-aware, trustworthy artificial intelligence: A paradigm shift for synthetic aperture radar,

    M. Datcu, Z. Huang, A. Anghel, J. Zhao, and R. Cacoveanu, “Explain- able, physics-aware, trustworthy artificial intelligence: A paradigm shift for synthetic aperture radar,” IEEE Geosci. Remote Sens. Mag. , vol. 11, no. 1, pp. 8–25, 2023

    2023

  43. [52]

    Physically explainable cnn for sar image classification,

    Z. Huang, X. Yao, Y . Liu, C. O. Dumitru, M. Datcu, and J. Han, “Physically explainable cnn for sar image classification,” ISPRS J. Photogramm. Remote Sens. , vol. 190, p. 25–37, Aug. 2022

    2022

  44. [53]

    The dual-band sar image fusion based foliage penetrating target detection method,

    H. Zhang, D. An, J. Li, L. Chen, D. Feng, Y . Song, and Z. Zhou, “The dual-band sar image fusion based foliage penetrating target detection method,” IEEE Trans. Geosci. Remote Sens. , 2024

    2024

  45. [54]

    A physically interpretable rice field extraction model for polsar imagery,

    J. Ge, H. Zhang, L. Xu, C. Sun, H. Duan, Z. Guo, and C. Wang, “A physically interpretable rice field extraction model for polsar imagery,” Remote Sens., vol. 15, no. 4, 2023

    2023

  46. [55]

    An entropy based classification scheme for land applications of polarimetric sar,

    S. Cloude and E. Pottier, “An entropy based classification scheme for land applications of polarimetric sar,”IEEE Trans. Geosci. Remote Sens., vol. 35, no. 1, pp. 68–78, 1997

    1997

  47. [56]

    Phenomenological theory of radar targets,

    J. R. Huynen, “Phenomenological theory of radar targets,” Ph.D. Dis- sertation, Delft University of Technology, Delft, Netherlands, 1970

    1970

  48. [57]

    A review of target decomposition theorems in radar polarimetry,

    S. Cloude and E. Pottier, “A review of target decomposition theorems in radar polarimetry,” IEEE Trans. Geosci. Remote Sens. , vol. 34, no. 2, pp. 498–518, 1996

    1996

  49. [58]

    Polsf: Polsar image datasets on san francisco,

    X. Liu, L. Jiao, F. Liu, D. Zhang, and X. Tang, “Polsf: Polsar image datasets on san francisco,” in Proc. Int. Conf. Intell. Sci.(ICIS) , 2022, pp. 214–219

    2022

  50. [59]

    Assessment of model-based polsar decompositions,

    T. L. Ainsworth, J.-S. Lee, and Y . Wang, “Assessment of model-based polsar decompositions,” in Proc. 2019 IEEE Int. Geosci. Remote Sens. Symp. (IGARSS), 2019, pp. 5019–5022

    2019

  51. [60]

    Polsar image classification using a hybrid complex- valued network (hybridcvnet),

    M. Q. Alkhatib, “Polsar image classification using a hybrid complex- valued network (hybridcvnet),”IEEE Geosci. Remote Sens. Lett., vol. 21, pp. 1–5, 2024

    2024

  52. [61]

    Cbam: Convolutional block attention module,

    S. Woo, J. Park, J.-Y . Lee, and I. S. Kweon, “Cbam: Convolutional block attention module,” in Proc. Eur. Conf. Comput. Vis. (ECCV) , 2018, pp. 3–19

    2018

  53. [62]

    Sf-cnn: A fast compression artifacts removal via spatial-to-frequency convolutional neural networks,

    T. Kim, H. Lee, H. Son, and S. Lee, “Sf-cnn: A fast compression artifacts removal via spatial-to-frequency convolutional neural networks,” in Proc. IEEE Int. Conf. Image Process. (ICIP) , 2019, pp. 3606–3610

    2019

  54. [63]

    Polsar image classification based on low- frequency and contour subbands-driven polarimetric senet,

    R. Qin, X. Fu, and P. Lang, “Polsar image classification based on low- frequency and contour subbands-driven polarimetric senet,” IEEE J. Sel. Topics Appl. Earth Observ. Remote Sens., vol. 13, pp. 4760–4773, 2020

    2020

  55. [64]

    A multichannel fusion convolutional neural network based on scattering mechanism for polsar image classification,

    Y . Wang, J. Cheng, Y . Zhou, F. Zhang, and Q. Yin, “A multichannel fusion convolutional neural network based on scattering mechanism for polsar image classification,” IEEE Geosci. Remote Sens. Lett. , vol. 19, pp. 1–5, 2022. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUS...

    2022

  56. [65]

    Local window attention transformer for polarimetric sar image classification,

    A. Jamali, S. K. Roy, A. Bhattacharya, and P. Ghamisi, “Local window attention transformer for polarimetric sar image classification,” IEEE Geosci. Remote Sens. Lett. , vol. 20, pp. 1–5, 2023

    2023

  57. [66]

    This looks like that: Deep learning for interpretable image recognition,

    C. Chen, O. Li, C. Tao, A. J. Barnett, J. Su, and C. Rudin, “This looks like that: Deep learning for interpretable image recognition,” arXiv:1806.10574, 2019

    2019 arXiv

  58. [67]

    Pol-insar-island - a benchmark dataset for multi-frequency pol-insar data land cover classification,

    S. Hochstuhl, N. Pfeffer, A. Thiele, S. Hinz, J. Amao-Oliva, R. Scheiber, A. Reigber, and H. Dirks, “Pol-insar-island - a benchmark dataset for multi-frequency pol-insar data land cover classification,” ISPRS J. Photogramm. Remote Sens. , vol. 10, p. 100047, 2023

    2023

  59. [68]

    Cnn-based polarimetric decomposition feature selection for polsar image classification,

    C. Yang, B. Hou, B. Ren, Y . Hu, and L. Jiao, “Cnn-based polarimetric decomposition feature selection for polsar image classification,” IEEE Trans. Geosci. Remote Sens. , vol. 57, no. 11, pp. 8796–8812, 2019. Jinqi Zhang received the B.E. degree in information and communicatio...

    2019

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.