ABSTRACT
To address the problems of low detection performance and large memory consumption of traditional smoke and fire detection methods in complex scenes such as grasslands. Based on the YOLOv5 model, a YOLOv5-GDE optimization model is proposed. The C3 module in YOLOv5 is replaced by GhostC3 with a smaller number of parameters, and some standard convolution blocks are replaced by depth-separable convolutions to make the model more lightweight. Finally, to solve the problem of unstable target regression frame, the EIoU loss function is introduced, which effectively improves the convergence speed and detection accuracy of the model. Experimental results on the homemade grassland smoke dataset show that the optimized model reduces the number of parameters and computational effort by 65.4% and 65.8%, respectively, compared with the original model, and the model size is only 36.8% of the original model, which is more suitable for smoke target detection in grassland scenes and more suitable for deployment in embedded devices with limited computational power, under the premise of ensuring detection accuracy.
- Editorial Department, LIU Y.2016 National grassland monitoring report[J]. China Animal Industry, 2017(08): 18-35.Google Scholar
- JIANG L, YU S, Wulantuya, Summary of grassland fire research[J]. Acta Agrestia Sinica,2018, 26(04): 791-803.Google Scholar
- CHEN J, ZHENG W, LIU C. Application of grassland fire monitoring based on Himawari-8 geostationary meteorological satellite data[J]. Journal of Natural Disasters,2017,26(04): 197-204.Google Scholar
- QIN X L, LI X T, LIU S C, Forest fire early warning and monitoring techniques using satellite remote sensing in China[J]. Journal of Remote Sensing,2020,24(05): 511-520.Google Scholar
- ZHU Y, JEON S, SUNG H, Developing UAV-based forest spatial information and evaluation technology for efficient forest management[J]. Sustainability, 2020, 12(23):10150.Google ScholarCross Ref
- ZHU M X, LIU Z Q, ZHANG X, Review of research on Video-based smoke detection algorithms[J/OL]. Computer Engineering and Applications, 2022,58(14):16-26.Google Scholar
- GIRSHICK R, DONAHUE J, DARRELL T, Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2014: 580-587.Google Scholar
- GIRSHICK R. Fast R-CNN[C]// Proceedings of the IEEE International Conference on Computer Vision, 2015: 1440-1448.Google Scholar
- REN S, HE K, GIRSHICK R, Faster R-CNN: towards real time object detection with region proposal net-works[C]//Advances in Annual Conference on Neural Information Processing Systems, 2015,28:91-99.Google Scholar
- LIU W, ANGUELOV D, ERHAN D, SSD: single shot ultibox detector[C]//European Conference on Computer Vision, 2016: 21-37.Google Scholar
- REDMON J, DIVVALA S, GIRSHICK R, You only look once: unified, real-time object detection[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition 2016: 779-788.Google Scholar
- REDMON J, FARHADI A. YOLO9000: better, faster, stronger[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017: 6517-6525.Google Scholar
- REDMON J, FARHADI A. YOLOv3: an Incremental Improvement[J]. arXiv: 1804.02767, 2018.Google Scholar
- BOCHKOVSKIY A, WANG C Y, LIAO H. YOLOv4: optimal speed and accuracy of object detection[J]. arXiv: 2004.10934, 2020.Google Scholar
- LI L, CAO L Y. Research on farmland flame and smoke image detection based on improved SSD algorithm[J]. Radio Engi-neering,2022,52(05): 783-789.Google Scholar
- XIE S H, ZHANG W Z, CHENG P, Firesmoke detection model based on YOLOv4 with channel attention[J]. Chinese Journal of Liquid Crystals and Displays,2021, 36(10): 1445-1453.Google Scholar
- ZHU T, YANG Z, ZHOU G X, A flame and smoke detection algorithm of the lightweight network[J]. Applied Science and Technology, 2022,49(02): 1-7.Google Scholar
- HAN K, WANG Y H, TIAN Q, GhostNet: more features from cheap operations[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020: 1577-1586.Google Scholar
- CHOLLET F. Xception: Deep learning with depthwise separable convolutions[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2017: 1800-1807.Google Scholar
- CAO Y K, GUI L A. Design of lightweight temporal convolutional network based on depthwise separable convolution[J]. Computer Engineering,2020,46(09): 95-100+109.Google Scholar
- REZATOFIGHI H, TSOI N, GWAK J, Generalized intersection over union: a metric and a loss for bounding box regression[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019: 658-666.Google Scholar
- ZHANG Y F, REN W, ZHANG Z, Focal and efficient IOU Loss for accurate bounding box regression[J]. arXiv: 2101.08158, 2021.Google Scholar
Index Terms
- Improved YOLOv5 lightweight grassland smoke detection algorithm
Recommendations
Forest Fire Detection Based on Improved YOLOv5
AIPR '21: Proceedings of the 2021 4th International Conference on Artificial Intelligence and Pattern RecognitionTraditional forest fire detection based on sensing devices has a small range, low sensitivity and high false alarm rate, while existing forest fire detection methods based on deep learning have poor identification effect and inaccurate identification of ...
An efficient lightweight CNN model for real-time fire smoke detection
AbstractEarly fire and smoke detection with computer vision have attracted much attention in recent years, and a lot of fire detectors based on deep neural network have been proposed to improve the detection accuracy. However, most current fire detectors ...
Adaptive estimation of visual smoke detection parameters based on spatial data and fire risk index
Standard wildfire smoke detection systems detect fires using remote cameras located at observation posts. Images from the cameras are analyzed using standard computer vision techniques, and human intervention is required only in situations in which the ...
Comments