Design and implementation of intelligent security inspection system on deep learning
Lu, Yuxiang (2026)
Kandidaatintutkielma
2026
School of Engineering Science, Tietotekniikka
Kaikki oikeudet pidätetään.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi-fe2026061268390
https://urn.fi/URN:NBN:fi-fe2026061268390
Tiivistelmä
This thesis presents an intelligent security inspection system designed for subway environments, with the goal of improving the accuracy and efficiency of prohibited-item detection in X-ray luggage images. In traditional security inspection, operators are required to continuously examine a large number of X-ray images, which can easily lead to fatigue and inconsistent detection performance during long working periods.
In this study, the YOLOv8 object detection model is used to identify and locate dangerous items in X-ray images. To improve the stability and detection capability of the model, a complete data processing workflow was developed, including image preprocessing, annotation conversion, and data augmentation. The model was trained and tested using the CLCXray Dataset dataset.
To support practical application, a lightweight client-server architecture was developed for image transmission, detection processing, and result display. Experimental results show that the proposed system can effectively detect prohibited items while maintaining relatively fast processing speed, demonstrating its potential value for intelligent subway security inspection systems.
In this study, the YOLOv8 object detection model is used to identify and locate dangerous items in X-ray images. To improve the stability and detection capability of the model, a complete data processing workflow was developed, including image preprocessing, annotation conversion, and data augmentation. The model was trained and tested using the CLCXray Dataset dataset.
To support practical application, a lightweight client-server architecture was developed for image transmission, detection processing, and result display. Experimental results show that the proposed system can effectively detect prohibited items while maintaining relatively fast processing speed, demonstrating its potential value for intelligent subway security inspection systems.