Design and Implementation of Indoor UAV Security System Integrating Deep Learning and Sensor

Проектирование и реализация системы безопасности на базе БПЛА для помещений с интеграцией глубокого обучения и датчиков
Lifeng Yang, Wenjie Yu, Yang Yue, Yidan Bian
2025-09-27

DWA obstacle avoidanceIR/ultrasonic sensorsSORT-EKF trackingYOLOv5-Liteindoor UAV security
In the context of the contemporary development of smart cities, traditional indoor surveillance systems, which rely on fixed cameras, encounter significant limitations, including a constrained sensing range and a substantial response delay. In this regard, the present paper proposes a novel indoor UAV security system, which is predicated on the use of deep learning and sensor technology. The system integrates key modules, including a lightweight target detection model (YOLOv5-Lite), a real-time multi-target tracking algorithm (SORT+EKF), and a path planning and multi-modal obstacle avoidance control (DWA+IR/Ultrasonic) algorithm. This integration facilitates high-precision recognition and realtime responsiveness to human behaviors within a complex indoor environment. In the context of the contemporary era of smart city development, traditional indoor surveillance systems that rely on fixed cameras encounter limitations, including constrained sensing range and considerable response delays. In this regard, the present paper puts forth a novel proposal for an indoor UAV security system that is predicated on the use of deep learning and sensor technology. The system integrates key modules, including a lightweight target detection model (YOLOv5-Lite), a real-time multi-target tracking algorithm (SORT + EKF), and a path planning and multi-modal obstacle avoidance control module (DWA+IR/Ultrasonic). This integration facilitates high-precision recognition and real-time responsiveness to people's behaviors in complex indoor environments.
1
DWA integrated with infrared and ultrasonic sensors supports path planning and real-time obstacle avoidance indoors.
2
SORT combined with an extended Kalman filter enables real-time tracking of multiple targets.
3
The integrated system is designed to overcome fixed-camera surveillance limitations, particularly constrained sensing range and response delays.
4
The paper proposes an indoor UAV security system combining deep learning, multi-target tracking, path planning, and multimodal obstacle avoidance.
5
YOLOv5-Lite provides a lightweight target-detection module for recognizing human behaviors in complex indoor environments.

the indoor UAV security system operating in complex indoor environments

high-precision human-behavior recognition, real-time multi-target tracking, and responsive navigation with obstacle avoidance

Publication Details
Publication Date
2025-09-27
Journal
Publisher
ISSN
Cited by
0
Access Type
Author Information
Authors
Lifeng Yang
Wenjie Yu
Yang Yue
Yidan Bian
Explore further
Open the scid.ai AI chat with a ready-made request: it will find papers on a similar topic and help build a literature review.
Find similar papers in the chat →
Make a presentation
100%