基于注意力机制的3D卷积双波段烟火识别方法

消防科学与技术 ›› 2023, Vol. 42 ›› Issue (2): 248-252.

基于注意力机制的3D卷积双波段烟火识别方法

宋俊猛1，殷松峰2，刘成2，米文忠2

（1.安徽建筑大学电子与信息工程学院，安徽合肥 230009；2.清华大学合肥公共安全研究院，安徽合肥 230601）

出版日期:2023-02-15 发布日期:2023-02-17
作者简介:作者简介：宋俊猛（1996- ），男，安徽阜阳人，安徽建筑大学计算机技术研究生，主要从事深度学习、烟火识别方面的研究，安徽省合肥市蜀山区紫云路292号，230009。
基金资助:
应急管理部消防救援局科技计划项目（2020XFCX30）

3D convolutional dual-band smoke and fire recognition method based on attention mechanism

Song Junmeng1, Yin Songfeng2, Liu Cheng2,Mi Wenzhong2

(1. School?of?Electronic?and?Information?Engineering, Anhui Jianzhu University, Anhui Hefei 230009, China; 2. Hefei Institute for Public Safety Research, Tsinghua University, Anhui Hefei 230601, China)

Online:2023-02-15 Published:2023-02-17

摘要/Abstract

摘要： 为提高复杂环境下烟火识别的精度，提出一种基于3D卷积和时空注意力机制的双波段烟火识别方法，该方法融合近红外和可见光双波段图像数据，使用视频流中基于时间的动态特征和基于空间的静态特征降低漏报率、误报率。实验结果表明，该算法在双波段数据集上的烟火识别精度达到99.90%，优于其他基于3D卷积的烟火识别算法，同时，模型具有较小的参数量，能够满足实时推理需求。因此，使用双波段特征的同时，结合注意力机制充分利用视频的动态信息，可以有效提高烟火识别精度。

Abstract: To improve the accuracy of smoke and fire recognition in complex environments, this paper proposes a dual-band smoke and fire recognition method based on 3D convolution and a spatio-temporal attention mechanism. The method in this paper fuses near-infrared and visible dual-band image data and uses time-based dynamic features and spatial-based static features in the video stream to reduce missed and false alarms. Experimental results show that the algorithm in this paper achieves 99.90% smoke and fire recognition accuracy on the dual-band data set, which is better than other 3D convolution-based smoke and fire recognition algorithms, while the model has a small number of parameters and can meet the real-time inference requirements. Therefore, the use of dual-band features combined with an attention mechanism to make full use of the dynamic information in the video can effectively improve the smoke and fire recognition accuracy.

Key words: smoke and flame recognition, 3D convolution, video dynamic features, multi-spectrum

宋俊猛, 殷松峰, 刘成, 米文忠. 基于注意力机制的3D卷积双波段烟火识别方法[J]. 消防科学与技术, 2023, 42(2): 248-252.

Song Junmeng, Yin Songfeng, Liu Cheng, Mi Wenzhong. 3D convolutional dual-band smoke and fire recognition method based on attention mechanism[J]. Fire Science and Technology, 2023, 42(2): 248-252.