Transfer Learning-Based Feature Fusion Model for Classification of Maneuver Weapon Systems
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
Convolutional neural network-based deep learning technology is the most commonly used in image identification,but it requires large-scale data for training. Therefore, application in specific fields in which data acquisitionis limited, such as in the military, may be challenging. In particular, the identification of ground weaponsystems is a very important mission, and high identification accuracy is required. Accordingly, various studieshave been conducted to achieve high performance using small-scale data. Among them, the ensemble method,which achieves excellent performance through the prediction average of the pre-trained models, is the mostrepresentative method; however, it requires considerable time and effort to find the optimal combination ofensemble models. In addition, there is a performance limitation in the prediction results obtained by using anensemble method. Furthermore, it is difficult to obtain the ensemble effect using models with imbalancedclassification accuracies. In this paper, we propose a transfer learning-based feature fusion technique for heterogeneousmodels that extracts and fuses features of pre-trained heterogeneous models and finally, fine-tuneshyperparameters of the fully connected layer to improve the classification accuracy. The experimental resultsof this study indicate that it is possible to overcome the limitations of the existing ensemble methods byimproving the classification accuracy through feature fusion between heterogeneous models based on transferlearning.
Abstract
Convolutional neural network-based deep learning technology is the most commonly used in image identification,but it requires large-scale data for training. Therefore, application in specific fields in which data acquisitionis limited, such as in the military, may be challenging. In particular, the identification of ground weaponsystems is a very important mission, and high identification accuracy is required. Accordingly, various studieshave been conducted to achieve high performance using small-scale data. Among them, the ensemble method,which achieves excellent performance through the prediction average of the pre-trained models, is the mostrepresentative method; however, it requires considerable time and effort to find the optimal combination ofensemble models. In addition, there is a performance limitation in the prediction results obtained by using anensemble method. Furthermore, it is difficult to obtain the ensemble effect using models with imbalancedclassification accuracies. In this paper, we propose a transfer learning-based feature fusion technique for heterogeneousmodels that extracts and fuses features of pre-trained heterogeneous models and finally, fine-tuneshyperparameters of the fully connected layer to improve the classification accuracy. The experimental resultsof this study indicate that it is possible to overcome the limitations of the existing ensemble methods byimproving the classification accuracy through feature fusion between heterogeneous models based on transferlearning.
참고문헌 (25)
- Sushil Singh, "Machine learning based distributed big data analysis framework for next generation web in IoT", Computer Science and Information Systems, 18(2), pp.597~618 (2021), National Library of Serbia[DOI]
- Bing Li, "Radiation Pneumonitis Prediction Using Multi-Omics Fusion Based on a Novel Machine Learning Pipeline", Human-centric Computing and Information Sciences, 12(0), pp.660~675 (2022), KIPS-CSWRG[DOI]
- Endah Kristiani, "Binary and Multi-Class Assessment of Face Mask Classification on Edge AI Using CNN and Transfer Learning", Human-centric Computing and Information Sciences, 12(0), pp.729~744 (2022), KIPS-CSWRG[DOI]
- 이한수, "전이학습 기반의 합성곱 신경망을 이용한 다중클래스 분류에 관한 연구", 한국지능시스템학회 논문지, 28(6), pp.531~537 (2018), 한국지능시스템학회
- 김용준, "딥 러닝 기법을 이용한 레이더 신호 분류 모델 연구", 한국군사과학기술학회지, 22(2), pp.170~178 (2019), 한국군사과학기술학회[DOI]
- 홍석준, "기계학습 방법을 이용한 레이더 신호 분류", 전기전자학회논문지, 22(1), pp.162~167 (2018), 한국전기전자학회[DOI]
- Hari Surrisyad, "A Fast Military Object Recognition using Extreme Learning Approach on CNN", International Journal of Advanced Computer Science and Applications, 11(12), pp.210~220 (2020), The Science and Information Organization[DOI]
- "ImageNet overview" (2020)[URL]
- 김귀훈, "딥러닝 기반의 이종 선생 네트워크를 지원하는 주요 파라미터 최적화 흐름정보 전이학습 기술", 한국정보기술학회논문지, 18(3), pp.21~29 (2020), 한국정보기술학회[DOI]
- 정미애, "컨볼루션 신경망의 앙상블 모델을 활용한 마스트 영상 기반 잠수함 탐지율 향상에 관한 연구", 한국군사과학기술학회지, 23(2), pp.115~124 (2020), 한국군사과학기술학회
- 황진용, "기동무기체계 분류를 위한 전이학습 기반 앙상블 모델 연구", 한국정보기술학회논문지, 19(12), pp.1~10 (2021), 한국정보기술학회[DOI]
- R. O. Lane, "Object detection in EO", IR and SAR images using low-SWAP hardware, 2021, 1~5
- Zhi Yang, "Deep transfer learning for military object recognition under small training set condition", Neural Computing and Applications, 31(10), pp.6469~6478 (2018), Springer Science and Business Media LLC[DOI]
- D. Legendre, "Information and Software Technologies", pp.220~242 (2020), Springer
- T. Hiippala, "Recognizing military vehicles in social media images using deep learning", Proceedings of 2017 IEEE International Conference on Intelligence and Security Informatics (ISI), pp.60~65 (2017)
- Yan Ouyang, "Military Vehicle Object Detection Based on Hierarchical Feature Representation and Refined Localization", IEEE Access, 10, pp.99897~99908 (2022), Institute of Electrical and Electronics Engineers (IEEE)[DOI]
- 정소영, "CNN의 깊은 특징과 전이학습을 사용한 보행자 분류", 인터넷정보학회논문지, 20(4), pp.91~102 (2019), 한국인터넷정보학회[DOI]
- K. Simonyan, "Very deep convolutional networks for large-scale image recognition", Proceedings of the 3rd International Conference on Learning Representations (ICLR), pp.1~14 (2014)
- M. Sandler, "MobileNetv2: inverted residuals and linear bottlenecks", Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp.4510~4520 (2018)
- K. He, "Deep residual learning for image recognition", Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp.770~778 (2016)
- G. Huang, "Densely connected convolutional networks", Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp.2261~2269 (2017)
- D. A. Clevert, "Fast and accurate deep network learning by exponential linear units (ELUs)", Proceedings of the 4th International Conference on Learning Representations (2016)
- N. Srivastava, "Dropout: a simple way to prevent neural networks from overfitting", The Journal of Machine Learning Research, 15(1), pp.1929~1958 (2014)
- D. P. Kingma, "Adam: a method for stochastic optimization", Proceedings of the 3rd International Conference on Learning Representations (ICLR) (2015)
- 박성욱, "앙상블 학습 알고리즘을 이용한 컨벌루션 신경망의 분류 성능 분석에 관한 연구", 멀티미디어학회논문지, 22(6), pp.665~675 (2019), 한국멀티미디어학회[DOI]