摘要(Abstract):
针对炼钢生产中采用人工经验判断转炉吹炼时期准确率较低的问题,利用某钢厂转炉炉口火焰图像数据,提出一种基于深度学习的改进DenseNet网络转炉吹炼时期识别算法;该算法以DenseNet-121网络结构为基础进行裁剪,同时在分类层引入Center损失函数,并在100次模型训练中选取精度较高、拟合性较好的一次进行测试。结果表明:该算法通过特征复用,保证了分类精度,裁剪后的网络结构能够提升运算速度;在分类层引入Center损失函数能够改进相邻转炉吹炼时期分类模糊的情况,缩短了平均识别时间,分类的平均精度提高至91.75%。
关键词(KeyWords): 深度学习;转炉吹炼;图像识别;DenseNet网络;Center损失函数
基金项目(Foundation): 国家自然科学基金项目(61763039)
作者(Author): 李爱莲,李晨筝,王懿喆,崔桂梅,解韶峰
DOI: 10.13349/j.cnki.jdxbn.20220223.001
参考文献(References):
[1] MOTTA D,SANTOS A á B,WINKLER I,et al.Application of convolutional neural networks for classification of adult mosquitoes in the field[J].PLoS One,2019,14(1):e0210829.
[2] MURTHY V,LOU L,SAMARAS D,et al.Center-focusing multi-task CNN with injected features for classification of glioma nuclear images[C]//2017 IEEE Winter Conference on Applications of Computer Vision (WACV),March 24-31,2017,Santa Rosa,USA.New York:IEEE,2017:834-841.
[3] TANG P J,WANG H L,KWONG S.G-MS2F:GoogleNet based multi-stage feature fusion of deep CNN for scene recognition[J].Neurocomputing,2017,225:188-197.
[4] SERMANET P,CHINTALA S,LECUN Y.Convolutional neural networks applied to house numbers digit classification[C]//Proceedings of the 21st International Conference on Pattern Recognition (ICPR2012),November 11-15,2012,Tsukuba,Japan.New York:IEEE,2012:3288-3291.
[5] TRIPATHI K,GUPTA A K,VYAS R G.Deep residual learning for image classification using cross validation[J].International Journal of Innovative Technology and Exploring Engineering,2020,9(6):1525-1530.
[6] GUBBI J,MARUSIC S,PALANISWAMI M.Smoke detection in video using wavelets and support vector machines[J].Fire Safety Journal,2009,44(8):1110-1115.
[7] 钟玲,张兴坤.基于SVM的视频图像火焰检测[J].软件工程,2017,20(6):1-4.
[8] 杨其睿.基于改进的DenseNet深度网络火灾图像识别算法[J].计算机应用与软件,2019,36(2):258-263.
[9] 李鹏举,刘辉,王彬,等.基于火焰彩色纹理复杂度特征的转炉炼钢吹炼状态识别[J].计算机应用,2015,35(1):283-288.
[10] 代照坤.面向转炉炼钢碳温预报模型的输入参量特征选择方法研究[D].昆明:昆明理工大学,2018.
[11] 赵多祯.基于图像处理的转炉终点预测研究[D].包头:内蒙古科技大学,2020.
[12] 汪宙.转炉冶炼中高碳钢过程及终点控制研究[D].北京:北京科技大学,2016.