XAI-based Hybrid CNN-LSTM Model for Forecasting and Management of Military Supply Chains with SAR Image Data

View at publisher Summarise with VIRAT
Authors: Omkar Akshara, Kumbala Pradeep Reddy, Kanukula Srujanasri, Ganga Sanuvala, A. Prakash
Cited by: 0
References: 10

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

The increase in relevance of Synthetic Aperture Radar (SAR) images in the past few years to include the military and its usage to observe, control and manage important elements in the military sector such as weapons, vehicles, food, fuel, and equipment has led to much research being focused on how to analyze and effectively use these types of images. To efficiently track, transport, store or deliver these types of resources for mission success and operational security are goals of the military supply chain management. This study presents a deep learning-based hybrid CNN2D-LSTM model that will classify SAR imagery with the intention of enhancing route prediction in military supply chains. The CNN2D portion of the network will be responsible for the spatial feature extraction of the SAR images while the LSTM portion will model sequential correlations of data points to increase prediction accuracy in dynamic environments. This study also incorporates Explainable AI (XAIs) best practices into the CNN2D-LSTM blended model through the use of LIME techniques to provide users with insight into the spatial areas that have contributed to the images being classified as seen in the experiment results. In conclusion, this study shows that the CNN2D and LSTM framework achieves excellent classification performance for the final completed training period and LIME should be viewed as an appropriate method for replicating SAR image classification findings while providing users with meaningful insight into SAR image classifications.

Fields of Study & Subjects

Advanced SAR Imaging Techniques Advanced Neural Network Applications Remote-Sensing Image Classification

Publication details

Journal / source
International Conference on Image Processing and Capsule Networks (ICIPCN)
Pages
283–288
Publication date
27 January 2026
Language
EN