XDMOM: A Real-Time Moving Object Detection System Based on a Dual-Spectrum Camera.

Area of Science:

• Computer Vision
• Embedded Systems
• Surveillance Technology

Background:

• Traditional video surveillance systems often struggle with power efficiency and outdoor performance.
• Real-time moving object detection in diverse environmental conditions remains a challenge.

Purpose of the Study:

• To develop a low-cost, power-efficient video surveillance system (XDMOM) for reliable outdoor moving object detection.
• To achieve all-day, 360-degree monitoring capabilities.
• To minimize false alarms through advanced processing techniques.

Main Methods:

• The XDMOM system integrates a dual-spectrum camera, a rotary platform, and a power-efficient NVIDIA GeForce GT1030 processor.
• Object detection is performed using the YOLOv4-tiny neural network.
• An adaptive weighted moving pipeline filter is employed to reduce false alarms in the time domain.

Main Results:

• The system achieves a correct alarm rate of 85.17% during the day and 81.79% at night for human detection in outdoor environments.
• The system's power consumption is maintained at a low level of 60-70 W.
• The system demonstrates superior performance compared to state-of-the-art video surveillance systems.

Conclusions:

• The developed XDMOM system offers a viable solution for low-cost, power-efficient, and high-performance outdoor video surveillance.
• The combination of hardware and software, including the YOLOv4-tiny network and filtering techniques, effectively addresses real-world detection challenges.
• The system's portability, enabled by a lithium battery, expands its applicability in various remote locations.