Radar HRRP Recognition Model Based on Pyramid Large Kernel Convolutional Attention and Stacked Transformer Encoders

Facing the challenge of radar high-resolution range profiles (HRRP) target recognition tasks with many target categories and limited training samples, existing radar HRRP target recognition models struggle to effectively capture and delineate the decision boundaries in the feature space of radar signals. Our paper proposes a radar HRRP recognition model based on pyramid large kernel convolutional attention (PLKC-Attention) and stacked transformer encoders (STE). The pyramid large kernel convolutional attention module captures multi-scale features from local to global levels, effectively extracting the details and structural characteristics of various HRRP target samples in a few-shot scenario. The stacked transformer encoders module constructs a deep encoders network that iteratively synthesizes the relationships between low-level details and structural characteristics, progressively resolving the high-level features corresponding to each class of targets, thereby enhancing the model's discrimination ability in scenarios with minor target class differences. Experimental results demonstrate that compared to other deep learning models, the model proposed in this paper exhibits superior recognition balance and accuracy in radar HRRP target recognition.