基于改进YOLOv8的室内火情检测

Indoor Fire Detection Based on Improved YOLOv8

摘要: 随着城市化进程的加快，室内火灾预警变得尤为重要。基于视觉的深度学习技术已成为火灾检测的研究热点，但在性能上仍有较大的改进空间。因此，本文设计了一种改良版的YOLOv8s算法，以提高室内火灾检测的准确性和实时性。改进的YOLOv8s算法通过整合Ghost模块和卷积块注意力机制(CBAM)，大幅降低了计算复杂度并增强了特征融合的作用。实验结果表明，该改进模型在模型参数量方面减少了44%，同时帧率提升了19.6%，检测精度也增加了2个百分点；在与其他主流算法进行对比时，在模型的精度、召回率和参数等均体现出均衡的优势。本文详细评估了这些改进对模型检测性能的影响，结果表明，改进后的YOLOv8s算法在检测速度和准确性上均表现出显著优势。本研究不仅为室内火灾监测提供了更高效的解决方案，也展示了深度学习在火灾检测中的广阔应用前景。

Abstract: With the acceleration of urbanization, indoor fire detection has become increasingly important. Visual-based deep learning technologies have emerged as a hot research topic in fire detection, but there is still significant room for improvement in performance. Therefore, we propose an improved version of the YOLOv8s algorithm to enhance the accuracy and real-time performance of indoor fire detection. The improved YOLOv8s algorithm integrates the Ghost module and Convolutional Block Attention Module (CBAM), significantly reducing computational complexity and enhancing feature fusion. Experimental results show that the improved model reduces the number of model parameters by 44%, while increasing the frame rate by 19.6% and improving detection accuracy by 2 percentage points. When compared with other mainstream algorithms, the improved model demonstrates balanced advantages in accuracy, recall rate, and parameters. We comprehensively evaluate the impact of these improvements on model detection performance, and the results show that the improved YOLOv8s algorithm exhibits significant advantages in both detection speed and accuracy. This study not only provides a more efficient solution for indoor fire detection but also demonstrates the broad application prospects of deep learning in fire detection.