Dataset Date released Public or not No of classes Classes

MSTAR 1990 public 10 BMP2, BTR70, T72,

T62, BDRM2, BTR60,
ZSU23/4, D7, ZIL131,
2S1

Paper Methodology Dataset date Ai Model Accuracy F1 score Findings Die score
Sparse Compared MSTAR 2016 SRC 75.28 Not mentioned SRC performs well on the MSTAR Not
Representation- performance of (Avg of 7 dataset, but overfitting and Mentioned
Based SAR Image SRC after methods) inadequate feature extraction
Target Classification preprocessing necessitate future
on the 10-Class dataset using improvements in robustness and
MSTAR Data Set PCA and feature selection
Downsampling

Synthetic aperture MSTAR 2024 JSR 99.8 Not mentioned the proposed multiview SAR Not mention
radar target 1. Image were target recognition method
recognition based on clustered on outperformed state-of-the-art
adaptive decision basis of techniques, demonstrating
fusion of multiple depression superior robustness and
views angles accuracy across various
i.e 30` and 45` operational conditions and view
configurations.
2.JSR was used
to classify
clustered images

3.errors are
converted to
posterior
possbilities

4. adapted
weights are
calculated on
basis of errors
based on shanon
entropy
 5. fused
posterior
probabilities are
obtained to
determine the
target label
Synthetic Aperture 1.Resnet with MSTAR 2023 Resnet 1)100% 1)100% The proposed attention 1
Radar Automatic simple attention 2)99.8% 2)99% mechanism enhances ResNet
Target Recognition mechanism architecture, achieving 100%
Based on a Simple 2. Resnet with identification accuracy for ten
Attention Mechanism one-policy different MSTAR SAR images,
learning rate demonstrating its effectiveness
for SAR target recognition
systems.
New SAR target a CNN MSTAR 2022 YOLO 99.7% 98.8% Proposed a new model SARON 1
recognition based on architecture with This model significantly
YOLO and very deep 19 weight layers, enhances SAR image target
multi-canonical utilizing small classification by adaptively
correlation analysis 3×3 fusing deep CNN features,
convolutional reducing noise impact, and
filters and achieving superior accuracy on
rectified linear the MSTAR dataset compared to
unit (ReLU) state-of-the-art methods
activations, while
omitting local
response
normalization to
enhance
computational
efficiency. The
model integrates
a linear support
vector machine
(L2-SVM) as a
baseline
classifier,
achieving state-
of-the-art
accuracy in SAR
 target
classification
Target Recognition of Euclidean MSTAR 2019 Templat 1)98.81 for __ proposed SAR ATR method __
Synthetic Aperture distance e 3 class shows superior performance in
Radar Images Based transform on method problem target recognition,
on Matching and target region demonstrating robustness
Similarity Evaluation residuals to 2)97.6 for against noise, occlusion, and
Between Binary enhance 10 class training data limitations.
Regions discrimination for problem
SAR ATR,
achieving robust
classification
under varying
conditions
Deep feature MSTAR 2017 KNN 1)98.81 for The method achieves higher
extraction and proposed LOGDET 3 class __ classification accuracy and __
combination for method uses VGG S1 problem outperforms state-of-the-art
synthetic aperture VGG-S1 to techniques on the MSTAR
radar target extract deep 2)97.6 for dataset
classification features from 10 class
SAR images, problem
applies
discriminant
correlation
analysis (DCA)
for feature
fusion, and
classifies targets
using K-NN with
LogDet
divergence-
based metric
learning.
Synthetic Aperture lightweight SAR MSTAR 2021 CNN 1)99.42 for The model significantly lowers
Radar SAR Image image with SOC __ parameter count and complexity __
Target Recognition recognition hybrid while maintaining high accuracy
Algorithm Based on model uses a attention 2)97.46 for and robustness in noisy SAR and
Attention Mechanism dual-channel mechnis eoc optical datasets.
convolution and m
hybrid attention
 mechanism to
reduce noise and
computational
complexity,
enhancing
feature extraction
with efficient
resource use.

[1]

[2]

[3]

[4]

[5]

[6]

[7]
1. Song, H., et al., Sparse representation-based SAR image target classification on the 10-class MSTAR data set. Applied Sciences, 2016. 6(1): p. 26.
2. Liang, J., Synthetic aperture radar target recognition based on adaptive decision fusion of multiple views. Journal of Electronic Imaging, 2024.
33(2): p. 023015-023015.
3. Ukwuoma, C.C.Z., Qin;Teninin, Bole W.;Yussif, Sophyani B.;Ejiyi, Chukwuebuka Joseph;Urama, Gilbert C.;Ukwuoma, Chibueze D.;Kind regards,
Ijeoma Little, Synthetic Aperture Radar Automatic Target Recognition Based on a Simple Attention Mechanism. International Journal of Interactive
Multimedia and Artificial Intelligence (IJIMAI), 2023.
4. Amrani, M., A. Bey, and A. Amamra, New SAR target recognition based on YOLO and very deep multi-canonical correlation analysis. International
Journal of Remote Sensing, 2022. 43(15-16): p. 5800-5819.
5. Ning, C., et al., Synthetic aperture radar target recognition using weighted multi-task kernel sparse representation. IEEE Access, 2019. 7: p.
181202-181212.
6. Amrani, M. and F. Jiang, Deep feature extraction and combination for synthetic aperture radar target classification. Journal of Applied Remote
Sensing, 2017. 11(4): p. 042616-042616.
7. Shi, B., et al., Synthetic aperture radar SAR image target recognition algorithm based on attention mechanism. IEEE Access, 2021. 9: p. 140512-
140524.