Please wait a minute...
浙江大学学报(工学版)
浙江大学学报(工学版)  2022, Vol. 56 Issue (10): 1891-1899    DOI: 10.3785/j.issn.1008-973X.2022.10.001
自动化技术、信息工程     
基于改进GhostNet-FCOS的火灾检测算法
张融(),张为*()
天津大学 微电子学院，天津 300072
Fire detection algorithm based on improved GhostNet-FCOS
Rong ZHANG(),Wei ZHANG*()
School of Microelectronics, Tianjin University, Tianjin 300072, China
 全文: PDF(2209 KB)   HTML
摘要：

针对现有火灾检测算法检测精度不佳和算法模型复杂度过高的问题，提出基于改进GhostNet-FCOS的火灾检测算法. 该算法以目标检测网络FCOS为基础，降低通道维数并选用GhostNet作为特征提取网络，以实现轻量化火灾检测算法. 引入动态卷积，在不增加网络宽度和深度的情况下优化主干网络的基础模块，提高对形态多变的火焰图像的特征提取能力. 增加空间注意力模块，优化网络空间特征的表达. 改进正负样本定义和回归损失函数，优化训练过程中算法模型对标注框内不同区域的关注程度. 在自建火灾数据集和公开数据集中的实验结果表明，该算法在检测精度和模型复杂度方面具有优势. 该算法在自建火灾数据集中的检测精度为90.9%，参数量为4.58×106，浮点计算量为31.45×109.
关键词： 火灾检测;  目标检测;  FCOS;  GhostNet;  动态卷积;  注意力模块    
Abstract:

A fire detection algorithm based on improved GhostNet-FCOS was proposed in view of the low detection accuracy and high complexity of existing fire detection algorithms. The algorithm was based on FCOS with reduced channel dimensions, and GhostNet was selected as the feature extraction network to implement a lightweight fire detection algorithm. Dynamic convolution was introduced to optimize the basic modules of the backbone without increasing width and depth, resulting in improved feature extraction ability for variable flames. A spatial attention module was introduced into the backbone network in order to optimize the expression of network spatial features. The definition of positive and negative samples and the regression loss function were improved to optimize the network’s attention to different areas in the ground truth box during the training process. The experimental results in self-built fire dataset and public dataset show that the algorithm has advantages in detection accuracy and model complexity. The detection accuracy of the algorithm in the self-built fire dataset was 90.9%, the amount of parameter was 4.58×106, and the floating point operation was 31.45×109.
Key words: fire detection    object detection    FCOS    GhostNet    dynamic convolution    attention module
收稿日期: 2021-11-03 出版日期: 2022-10-25
CLC:  TP 391  
基金资助: 新一代人工智能科技重大专项资助项目(19ZXZNGX00030)
通讯作者: 张为     E-mail: rongzhang@tju.edu.cn;tjuzhangwei@tju.edu.cn
作者简介: 张融（1997—），男，硕士生，从事数字图像处理、模式识别的研究. orcid.org/0000-0002-9092-6426. E-mail： rongzhang@tju.edu.cn
服务　　
把本文推荐给朋友
加入引用管理器
E-mail Alert
作者相关文章　　
张融
张为

引用本文:

张融,张为. 基于改进GhostNet-FCOS的火灾检测算法[J]. 浙江大学学报(工学版), 2022, 56(10): 1891-1899.

Rong ZHANG,Wei ZHANG. Fire detection algorithm based on improved GhostNet-FCOS. Journal of ZheJiang University (Engineering Science), 2022, 56(10): 1891-1899.

链接本文:

https://www.zjujournals.com/eng/CN/10.3785/j.issn.1008-973X.2022.10.001    或    https://www.zjujournals.com/eng/CN/Y2022/V56/I10/1891

图 1  基于改进GhostNet-FCOS的火灾检测算法的整体结构
输入 操作 扩张系数 输出 注意力 步长
3 3×3 conv — 16 — 2
16 Ghost bottleneck 16 16 — 1
16 Ghost bottleneck 48 24 — 2
24 Ghost bottleneck 72 24 — 1
24 Ghost bottleneck 72 40 √ 2
40 Ghost bottleneck 120 40 √ 1
40 Ghost bottleneck 240 80 — 2
80 Ghost bottleneck 200 80 — 1
80 Ghost bottleneck 184 80 — 1
80 Ghost bottleneck 184 80 — 1
80 Ghost bottleneck 480 112 √ 1
112 Ghost bottleneck 672 112 √ 1
112 Ghost bottleneck 672 160 √ 2
160 Ghost bottleneck 960 160 — 1
160 Ghost bottleneck 960 160 √ 1
160 Ghost bottleneck 960 160 — 1
160 Ghost bottleneck 960 160 √ 1
160 1×1 conv — 960 — 1
表 1  改进GhostNet的网络结构
图 2  结合动态卷积的Ghost模块
图 3  空间注意力模块
图 4  中心采样策略
图 5  火灾检测公共数据集
检测网络-通道数 主干网络 AP50/% AR/% P/106 FLOPs/109 v/(帧·s?1)
FCOS-256 ResNet50 91.4 56.2 74.13 200.5 18.0
FCOS-128 ShuffleNetV2 1.0× (g=3) 83.9 51.7 3.06 31.99 33.5
FCOS-128 ShuffleNetV2 2.0× 85.6 53.9 7.34 41.68 32.3
FCOS-128 MobileNetV2 86.2 53.2 4.04 35.22 29.1
FCOS-128 MobileNetV3 86.6 52.5 4.74 33.35 29.6
FCOS-128 GhostNet 86.5 52.4 4.45 30.45 28.6
表 2  轻量级主干网络的对比
算法 动态卷积模块 空间注意力模块 监督信号和损失函数 AP50/% AR/% P/106 v/(帧·s?1)
FCOS — — — 91.4 56.2 74.13 18.0
GhostNet-FCOS — — — 86.5 52.4 4.45 28.6
本文算法 √ — — 88.7 53.8 4.58 25.1
本文算法 √ √ — 90.1 54.2 4.58 24.1
本文算法 √ √ √ 90.9 55.1 4.58 24.1
表 3  消融实验的结果
图 6  改进监督信号和损失函数前、后的检测效果对比
算法 TPR/% TNR/%
V1 V2 V3 V4 V5
文献[25] 92.50 94.21 — 95.45 —
文献[26] 92.25 94.59 — 96.15 —
文献[27] 93.00 94.98 — 96.50 —
文献[28] 93.10 95.70 — 100 —
文献[2] 96.15 100.0 94.29 100 95.20
文献[29] — 91.20 — 95.80 —
文献[3] 93.30 — — — 94.70
文献[30] 91.00 93.00 — 100 —
本文算法1 96.66 96.53 99.59 95.41 96.10
本文算法2 98.75 100 100 99.23 99.35
表 4  与基于传统图像特征的火灾检测算法对比
图 7  Ghost-FCOS在改进前、后的检测效果对比
算法 AP50/% AR/% P/106 FLOPs/109
YOLOV3 Tiny [31] 76.5 43.2 8.67 16.24
YOLOV3-MobileNet[32] 82.6 46.8 22.67 53.46
SSD-MobileNetV2 [33] 76.6 46.9 5.45 8.08
本文算法 90.9 55.1 4.58 31.45
表 5  与用于火灾检测的轻量级目标检测网络对比
图 8  实验与真实火灾的检测效果图
1 FOGGIA P, SAGGESE A, VENTO M Real-time fire detection for video-surveillance applications using a combination of experts based on color, shape, and motion[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2015, 25 (9): 1545- 1556
doi: 10.1109/TCSVT.2015.2392531
2 严云洋, 杜静, 高尚兵, 等 融合多特征的视频火焰检测[J]. 计算机辅助设计与图形学学报, 2015, 27 (3): 433- 440
YAN Yun-yang, DU Jing, GAO Shang-bing, et al Video flame detection based on fusion of multi-feature[J]. Journal of Computer-Aided Design and Computer Graphics, 2015, 27 (3): 433- 440
3 HASHEMZADEH M, ZADEMEHDI A Fire detection for video surveillance applications using ICA k-medoids-based color model and efficient spatio-temporal visual features[J]. Expert Systems with Applications, 2019, 130: 60- 78
doi: 10.1016/j.eswa.2019.04.019
4 MUHAMMAD K, AHMAD J, BAIK S W Early fire detection using convolutional neural networks during surveillance for effective disaster management[J]. Neurocomputing, 2018, 288: 30- 42
doi: 10.1016/j.neucom.2017.04.083
5 MUHAMMAD K, AHMAD J, MEHMOOD I, et al Convolutional neural networks based fire detection in surveillance videos[J]. IEEE Access, 2018, 6: 18174- 18183
doi: 10.1109/ACCESS.2018.2812835
6 MUHAMMAD K, AHMAD J, LV Z, et al Efficient deep CNN-based fire detection and localization in video surveillance applications[J]. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 2018, 49 (7): 1419- 1434
7 MUHAMMAD K, KHAN S, ELHOSENY M, et al Efficient fire detection for uncertain surveillance environment[J]. IEEE Transactions on Industrial Informatics, 2019, 15 (5): 3113- 3122
doi: 10.1109/TII.2019.2897594
8 LI S, YAN Q, LIU P An efficient fire detection method based on multiscale feature extraction, implicit deep supervision and channel attention mechanism[J]. IEEE Transactions on Image Processing, 2020, 29: 8467- 8475
doi: 10.1109/TIP.2020.3016431
9 JEON M, CHOI H S, LEE J, et al Multi-scale prediction for fire detection using convolutional neural network[J]. Fire Technology, 2021, 57 (5): 1- 19
10 CHAOXIA C Y, SHANG W W, ZHANG F Information-guided flame detection based on faster R-CNN[J]. IEEE Access, 2020, 8: 58923- 58932
doi: 10.1109/ACCESS.2020.2982994
11 晋耀, 张为 采用Anchor-Free网络结构的实时火灾检测算法[J]. 浙江大学学报: 工学版, 2020, 54 (12): 2430- 2436
JIN Yao, ZHANG Wei Real-time fire detection algorithm with anchor-free network architecture[J]. Journal of Zhejiang University: Engineering Science, 2020, 54 (12): 2430- 2436
12 李欣健, 张大胜, 孙利雷, 等 复杂场景下基于CNN的轻量火焰检测方法[J]. 模式识别与人工智能, 2021, 34 (5): 415- 422
LI Xin-jian, ZHANG Da-sheng, SUN Li-lei, et al CNN-based lightweight fire detection method in complex scenes[J]. Pattern Recognition and Artificial Intelligence, 2021, 34 (5): 415- 422
doi: 10.16451/j.cnki.issn1003-6059.202105004
13 KIM B, LEE J A video-based fire detection using deep learning models[J]. Applied Sciences, 2019, 9 (14): 2862
doi: 10.3390/app9142862
14 TIAN Z, SHEN C, CHEN H, et al. FCOS: fully convolutional one-stage object detection [C]// Proceedings of the IEEE International Conference on Computer Vision. Seoul: IEEE, 2019: 9627–9636.
15 HAN K, WANG Y, TIAN Q, et al. Ghostnet: more features from cheap operations [C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Seattle: IEEE, 2020: 1580-1589.
16 LIN T Y, DOLLAR P, GIRSHICK R, et al. Feature pyramid networks for object detection [C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Honolulu: IEEE, 2017: 2117–2125.
17 YANG B, BENDER G, LE Q V, et al CondConv: conditionally parameterized convolutions for efficient inference[J]. Advances in Neural Information Processing Systems, 2019, 32: 1307- 1318
18 HU J, SHEN L, ALBANIE S, et al Squeeze-and-excitation networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42 (8): 2011- 2023
doi: 10.1109/TPAMI.2019.2913372
19 WOO S, PARK J, LEE J Y, et al. CBAM: convolutional block attention module [C]// European Conference on Computer Vision. Munich: Springer, 2018: 1-19.
20 REZATOFIGHI H, TSOI N, GWAK J, et al. Generalized intersection over union: a metric and a loss for bounding box regression [C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Long Beach: IEEE, 2019: 658-666.
21 LIN T Y, MAIRE M, BELONGIE S, et al. Microsoft coco: common objects in context [C]// European Conference on Computer Vision. Zurich: Springer, 2014: 740-755.
22 MA N, ZHANG X, ZHENG H T, et al. Shufflenet v2: practical guidelines for efficient CNN architecture design [C]// Proceedings of the European Conference on Computer Vision. Munich: Springer, 2018: 116-131.
23 SANDLER M, HOWARD A, ZHU M, et al. Mobilenetv2: inverted residuals and linear bottlenecks [C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Salt Lake City: IEEE, 2018: 4510-4520.
24 HOWARD A, SANDLER M, CHEN B, et al. Searching for mobilenetv3 [C]// International Conference on Computer Vision. Seoul: IEEE, 2019: 1314-1324.
25 KO B C, CHEONG K H, NAM J Y Fire detection based on vision sensor and support vector machines[J]. Fire Safety Journal, 2009, 44 (3): 322- 329
doi: 10.1016/j.firesaf.2008.07.006
26 BORGES P V K, IZQUIERDO E A probabilistic approach for vision-based fire detection in videos[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2010, 20 (5): 721- 731
doi: 10.1109/TCSVT.2010.2045813
27 TRUONG T X, KIM J M Fire flame detection in video sequences using multi-stage pattern recognition techniques[J]. Engineering Applications of Artificial Intelligence, 2012, 25 (7): 1365- 1372
doi: 10.1016/j.engappai.2012.05.007
28 KONG S G, JIN D, LI S Fast fire flame detection in surveillance video using logistic regression and temporal smoothing[J]. Fire Safety Journal, 2016, 79: 37- 43
doi: 10.1016/j.firesaf.2015.11.015
29 吴茜茵, 严云洋, 杜静, 等 多特征融合的火焰检测算法[J]. 智能系统学报, 2015, 10 (2): 240- 247
WU Xi-yin, YAN Yun-yang, DU Jing, et al Fire detection based on fusion of multiple features[J]. CAAI Transactions on Intelligent Systems, 2015, 10 (2): 240- 247
doi: 10.3969/j.issn.1673-4785.201406022
30 梅建军, 张为 基于ViBe与机器学习的早期火灾监测算法[J]. 光学学报, 2018, 38 (7): 60- 67
MEI Jian-jun, ZHANG Wei Early fire detection algorithm based on ViBe and machine learning[J]. Acta Optica Sinica, 2018, 38 (7): 60- 67
31 KANG L W, WANG I S, CHOU K L, et al. Image-based real-time fire detection using deep learning with data augmentation for vision-based surveillance applications [C]// 16th IEEE International Conference on Advanced Video and Signal Based Surveillance. Taipei: IEEE, 2019: 1-4.
32 DAI Z. Image flame detection method based on improved yolov3 [C]// IOP Conference Series: Earth and Environmental Science. Xi’an: IOP Publishing, 2021, 693(1): 012012.
[1] 张娜,戚旭磊,包晓安,吴彪,涂小妹,金瑜婷. 基于优化预测定位的单阶段目标检测算法[J]. 浙江大学学报(工学版), 2022, 56(4): 783-794.
[2] 褚晶辉,史李栋,井佩光,吕卫. 适用于目标检测的上下文感知知识蒸馏网络[J]. 浙江大学学报(工学版), 2022, 56(3): 503-509.
[3] 周金海,周世镒,常阳,吴耿俊,王依川. 基于超宽带雷达基带信号的多人目标跟踪[J]. 浙江大学学报(工学版), 2021, 55(6): 1208-1214.
[4] 徐利锋,黄海帆,丁维龙,范玉雷. 基于改进DenseNet的水果小目标检测[J]. 浙江大学学报(工学版), 2021, 55(2): 377-385.
[5] 王浩远,梁煜,张为. 融合多分辨率表征的实时烟雾分割算法[J]. 浙江大学学报(工学版), 2021, 55(12): 2334-2341.
[6] 刘清清,周志勇,范国华,钱旭升,胡冀苏,陈光强,戴亚康. 基于3D scSE-UNet的肝脏CT图像半监督学习分割方法[J]. 浙江大学学报(工学版), 2021, 55(11): 2033-2044.
[7] 牛英杰,苏燕辰,程敦诚,廖家,赵海波,高永强. 高铁接触网U型抱箍螺母故障检测算法[J]. 浙江大学学报(工学版), 2021, 55(10): 1912-1921.
[8] 郑浦,白宏阳,李伟,郭宏伟. 复杂背景下的小目标检测算法[J]. 浙江大学学报(工学版), 2020, 54(9): 1777-1784.
[9] 张峻宁,苏群星,刘鹏远,王正军,谷宏强. 基于空间约束的自适应单目3D物体检测算法[J]. 浙江大学学报(工学版), 2020, 54(6): 1138-1146.
[10] 郑晨斌,张勇,胡杭,吴颖睿,黄广靖. 目标检测强化上下文模型[J]. 浙江大学学报(工学版), 2020, 54(3): 529-539.
[11] 晋耀,张为. 采用Anchor-Free网络结构的实时火灾检测算法[J]. 浙江大学学报(工学版), 2020, 54(12): 2430-2436.
[12] 林志洁,罗壮,赵磊,鲁东明. 特征金字塔多尺度全卷积目标检测算法[J]. 浙江大学学报(工学版), 2019, 53(3): 533-540.
[13] 叶芳芳,许力. 实时的静止目标与鬼影检测及判别方法[J]. 浙江大学学报(工学版), 2015, 49(1): 181-185.
[14] 刘辉涛,汪李明,李建龙. 声纳强脉冲干扰的自适应抵消方法[J]. J4, 2011, 45(3): 515-519.
[15] 刘士荣, 王凯, 邱雪娜. 基于自适应混合高斯模型全方位视觉目标检测[J]. J4, 2010, 44(7): 1387-1393.