Simultaneous Ship Detection and Orientation Estimation in SAR Images Based on Attention Module and Angle Regression

Area of Science:

• Remote Sensing
• Computer Vision
• Artificial Intelligence

Background:

• Ship detection and orientation estimation are crucial for marine surveillance.
• Sequential processing in previous methods is time-consuming and inefficient.

Purpose of the Study:

• To develop a unified approach for simultaneous ship detection and angle estimation in SAR images.
• To improve the efficiency and accuracy of marine surveillance systems.

Main Methods:

• Utilizing a convolutional neural network based on the Single Shot Detector (SSD) framework.
• Implementing rotatable bounding boxes for tighter object localization and orientation prediction.
• Employing a semantic aggregation method for feature fusion and an attention module for enhanced feature selection.
• Designing multi-orientation anchors and embedding angular regression for simultaneous prediction.

Main Results:

• The proposed method achieves simultaneous detection, classification, and angle estimation in a single forward pass.
• Rotatable bounding boxes provide more accurate localization and robustness to background noise.
• Semantic aggregation and attention mechanisms improve the detection of small ships.
• The integrated angular regression enhances prediction accuracy without significant computational overhead.

Conclusions:

• The novel approach effectively combines ship detection and angle estimation, outperforming traditional methods.
• The method offers a more efficient and accurate solution for marine surveillance using SAR imagery.
• Further validation on diverse datasets is recommended to confirm generalizability.