Convolutional Neural Networks for Automatic Identification of Individuals at Terrestrial Terminals
The objective of this study was to develop an automated system for the identification of wanted individuals in terrestrial terminals using Convolutional Neural Networks (CNN). The research was conducted under a quantitative approach and a quasi-experimental design. A private dataset, intended for ed...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10974742/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The objective of this study was to develop an automated system for the identification of wanted individuals in terrestrial terminals using Convolutional Neural Networks (CNN). The research was conducted under a quantitative approach and a quasi-experimental design. A private dataset, intended for educational purposes, consisting of images and videos of individuals in dynamic environments, was employed to enable identification testing under real-world conditions. The methodology encompassed the structured loading of biometric data, facial detection using the Multi-Task Cascaded Convolutional Neural Network (MTCNN) model, and the generation of facial embeddings through the InceptionResNetV1 model. Extracted features and data were stored in a MySQL database. To optimize the search process during real-time identification, the embeddings were transferred to FAISS, a library optimized for similarity search in large-scale datasets. In FAISS, the embeddings were stored in vector format to facilitate fast and efficient querying. Subsequently, identification tasks were performed on video sequences captured in real time. The results revealed high performance of the proposed system, achieving an 89% accuracy in facial detection and 98.60% accuracy in real-time person identification when compared against the data stored in the database. Finally, the trained models were integrated into a web application that enables real-time identification using IP cameras, leveraging YOLO version 8 architecture for tracking identified individuals. These results confirm that deep learning models can be effectively incorporated into surveillance and control systems in public spaces, enhancing existing security mechanisms and offering a viable solution for real-time automatic person identification. |
|---|---|
| ISSN: | 2169-3536 |