研究目的
Investigating the real-time detection of concealed metallic weapons on the human body using passive millimeter wave (PMMW) imagery based on the YOLOv3 algorithm.
研究成果
The YOLOv3-53 model demonstrates superior performance in detecting concealed weapons in PMMW images, achieving a high detection accuracy (95% mAP) and a real-time detection speed (36 FPS). The study highlights the potential of deep learning algorithms for real-time security applications, even with limited sample data.
研究不足
The study is limited by the quality and quantity of the PMMW image dataset. The detection accuracy may be affected by the thickness of clothing and the position of concealed objects. The computational complexity of the YOLOv3-53 model may limit its deployment in resource-constrained environments.
1：Experimental Design and Method Selection:
The study employs the YOLOv3 algorithm for real-time detection of concealed weapons in PMMW images. It involves training two models, YOLOv3-13 and YOLOv3-53, and comparing their performance with the SSD-VGG16 algorithm.
2：Sample Selection and Data Sources:
A dataset of 1634 PMMW images captured at the 34 GHz band from the SAIR-U system developed by Beihang University is used. The dataset includes images of humans carrying metal guns under different clothing thicknesses, temperatures, and imaging speeds.
3：List of Experimental Equipment and Materials:
The PMMW real-time imager SAIR-U, developed by the Microwave Laboratory of Beihang University, is used for data collection. The system operates at the Ka band (34 GHz) with a sensitivity to temperature of 1~2 K.
4：Experimental Procedures and Operational Workflow:
The YOLOv3 models are trained on the dataset with specific hyperparameters (batch_size: 8, momentum:
5：9, learning rate:
0.001, etc.). The models are evaluated based on detection accuracy, speed, and computational resources.
6：001, etc.). The models are evaluated based on detection accuracy, speed, and computational resources.
Data Analysis Methods:
5. Data Analysis Methods: The performance is evaluated using Intersection Over Union (IoU) and mean average precision (mAP) metrics. The models' detection speeds and accuracy are compared.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
