An Enhanced YOLOv8n-Based Method for Fire Detection in Complex Scenarios.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Environmental Monitoring

Background:

• Climate change is increasing fire frequency, necessitating advanced detection systems.
• Existing systems struggle with low visibility and dynamic disturbances.
• Public safety and ecological protection demand robust fire detection.

Purpose of the Study:

• To develop a comprehensive multi-module fire detection framework.
• To enhance fire detection accuracy and robustness in complex environments.
• To address data scarcity using generative adversarial networks.

Main Methods:

• Image enhancement and lightweight object detection using visual computing.
• Projected Generative Adversarial Network (Projected GAN) for data synthesis.
• Improved YOLOv8n architecture with BiFormer Attention, Agent Attention, and Compact Channel Compression (CCC) modules.

Main Results:

• Achieved high image restoration quality (PSNR up to 34.67 dB, SSIM up to 0.968).
• Reached significant detection performance (mAP of 0.858).
• Demonstrated superior performance over baseline methods on synthetic and real-world data.

Conclusions:

• The proposed framework offers a reliable and deployable solution for real-time fire monitoring.
• The system effectively handles low visibility and dynamic disturbances.
• This advancement contributes to improved early warning systems for fire events.