FDCT: Frequency-Aware Decomposition and Cross-Modal Token-Alignment for Multisensor Target Classification

In automatic target recognition (ATR) systems, sensors may fail to capture discriminative, fine-grained detail features due to environmental conditions, noise created by CMOS chips, occlusion, parallaxes, and sensor misalignment. Therefore, multisensor image fusion is an effective choice to overcome these constraints. However, multimodal image sensors are heterogeneous, and have domain and granularity gap. In addition, the multisensor images can be misaligned due to intricate background clutters, fluctuating illumination conditions, and uncontrolled sensor settings. In this article, to overcome these issues, we decompose, align, and fuse multiple image sensor data for target classification. We extract the domain-specific and domain-invariant features from each sensor data. We propose to develop a shared unified discrete token (UDT) space between sensors to reduce the domain and granularity gap. In addition, we develop an alignment module to overcome the misalignment between multisensors and emphasize the discriminative representation of the UDT space. In the alignment module, we introduce sparsity constraints to provide a better cross-modal representation of the UDT space and robustness against various sensor settings. We achieve superior classification performance compared to single-modality classifiers and several state-of- the-art multimodal fusion algorithms on four multisensor ATR datasets. Furthermore, the experimental results demonstrate that our multisensor classifier outperforms state-of-the-art single modality classifiers, e.g., EfficentNet, ViT, and ConvNeXt, in the VEDAI ATR dataset.