Related Experiment Video
Updated: Jul 23, 2025
03:31
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
Published on: December 15, 2023
581
Suspicious Object Detection for Millimeter-Wave Images With Multi-View Fusion Siamese Network.
Summary
This study introduces a new method for detecting suspicious objects in millimeter-wave (MMW) images using Siamese networks. The technique enhances accuracy by analyzing symmetrical body parts and fusing multi-view data.
Area of Science:
- Computer Vision
- Medical Imaging
- Security Technology
Background:
- Millimeter-wave (MMW) imaging is utilized in public security due to privacy and safety benefits.
- Detecting small, low-resolution, and reflection-weak suspicious objects in MMW images presents significant challenges.
Purpose of the Study:
- To develop a robust suspicious object detection model for MMW images.
- To address limitations of existing methods that require complete training datasets with accurate annotations.
Main Methods:
- A Siamese network approach integrating pose estimation and image segmentation is employed.
- Human images are segmented into symmetrical body parts to learn inter-part similarity.
- Multi-view MMW images are fused using decision-level and feature-level strategies with attention mechanisms to mitigate misdetections from restricted fields of view.
Main Results:
- Experimental results on measured MMW images demonstrate favorable detection accuracy.
- The proposed model exhibits effective performance in practical applications, showing good detection speed.
Conclusions:
- The developed Siamese network-based detector effectively addresses challenges in MMW suspicious object detection.
- The integration of pose estimation, segmentation, and multi-view fusion strategies proves effective for enhancing detection capabilities in real-world scenarios.
