Physics-Guided Detector for SAR Airplanes

Abstract

The disperse structure distributions (discreteness) and variant scattering characteristics (variability) of SAR airplane targets lead to special challenges of object detection and recognition. The current deep learning-based detectors encounter challenges in distinguishing fine-grained SAR airplanes against complex backgrounds. To address it, we propose a novel physics-guided detector (PGD) learning paradigm for SAR airplanes that comprehensively investigate their discreteness and variability to improve the detection performance. It is a general learning paradigm that can be extended to different existing deep learning-based detectors with "backbone-neck-head" architectures. The main contributions of PGD include the physics-guided self-supervised learning, feature enhancement, and instance perception, denoted as PGSSL, PGFE, and PGIP, respectively. PGSSL aims to construct a self-supervised learning task based on a wide range of SAR airplane targets that encodes the prior knowledge of various discrete structure distributions into the embedded space. Then, PGFE enhances the multi-scale feature representation of a detector, guided by the physics-aware information learned from PGSSL. PGIP is constructed at the detection head to learn the refined and dominant scattering point of each SAR airplane instance, thus alleviating the interference from the complex background. We propose two implementations, denoted as PGD and PGD-Lite, and apply them to various existing detectors with different backbones and detection heads. The experiments demonstrate the flexibility and effectiveness of the proposed PGD, which can improve existing detectors on SAR airplane detection with fine-grained classification task (an improvement of 3.1\% mAP most), and achieve the state-of-the-art performance (90.7\% mAP) on SAR-AIRcraft-1.0 dataset. The project is open-source at \url{https://github.com/XAI4SAR/PGD}.

Figures (8)

• Figure 1: (a) SAR ship and vehicle targets depict continuous scattering characteristics and regular shapes. (b) As a comparison, SAR airplanes are more discrete and variable.
• Figure 2: Left: The general deep learning-based detector architecture. Right: The proposed physics-guided detector learning paradigm. It consists of (1) physics-guided self-supervised learning (PGSSL) based on a surrogate task of airplane structure distribution prediction, (2) physics-guided feature enhancement (PGFE) to improve the multi-scale features of the detector's neck, and (3) physics-guided instance perception (PGIP) at the detector's head.
• Figure 3: The proposed physics-guided self-supervised learning (PGSSL) of SAR airplanes. A surrogate task of scattering structure distribution prediction is defined to construct the self-supervised learning. The proposed PGSSL can be realized with different implementations. E.g., (1) CNN based PGSSL can be designed to aggregate the multi-scale hierarchical features to represent the scattering structure distribution. (2) Transformer based PGSSL can be also designed to capture the long-range dependencies among local features of SAR airplane to represent the scattering distribution.
• Figure 4: The detailed implementation of the proposed physics-guided detector for SAR airplanes. With the input of $x$, the pre-trained PGSSL model is frozen to extract the physics-aware features $F_\mathrm{P}$ that would have representative activations in $x$.
• Figure 5: Visualization of the detection results on test data. (a) Ground Truth (b) DEQDet wang2023deep (c) YOLOv5 Jocher_YOLOv5_by_Ultralytics_2020 (d) YOLOv8 Jocher_Ultralytics_YOLO_2023 (e) PGD-YOLO