作者：曹君¹, 李爱莲¹, 解韶峰², 崔桂梅¹

作者单位：1. 内蒙古科技大学信息工程学院，内蒙古 包头 014010;
2. 内蒙古科技大学基建处，内蒙古 包头 014010

关键词：下渣检测;YOLOv5s;注意力机制;损失函数;BiFPN

摘要：

针对人工目测法、红外热像检测法在转炉下渣状态检测中检测精度和实时性较差带来挡渣操作不及时，进而影响钢成品质量的问题，提出一种基于改进YOLOv5s的转炉下渣状态检测方法。在主干网络融合卷积注意力（CBAM），增强算法特征提取能力；在颈部层引入加权双向特征金字塔结构（BiFPN），将主干结构的原始特征信息与输出节点的特征信息进行多层次融合，并给予不同特征相应的权重，获得更加丰富的特征图；在检测层使用EIoU Loss函数优化模型性能，提升预测框的收敛速度。实验结果表明：改进后模型的均值平均精度（mAP）达到91.8%，每秒传输帧数（FPS）为87.7 f/s，相比原模型分别提高4.6%和11.4%。

Research on the state detection algorithm of slag in converter based on improved YOLOv5s
CAO Jun¹, LI Ailian¹, XIE Shaofeng², CUI Guimei¹
1. School of Information Engineering, Inner Mongolia University of Science and Technology, Baotou 014010, China;
2. Department of Infrastructure, Inner Mongolia University of Science and Technology,Baotou 014010, China
Abstract: Aiming at the problem of untimely slag blocking operation affects the quality of steel products caused by poor detection accuracy and real-time performance of manual visual inspection method and infrared thermal imaging detection method in the detection of converter slag state, an improved YOLOv5s-based convert slag state detection method is proposed. Integrating the convolutional block attention module (CBAM) in the backbone network to enhance the feature extraction performance of the algorithm.The bidirectional feature pyramid network (BiFPN) is applied in the neck layer, the original feature information of the backbone structure and the feature information of the output nodes are multi-levelly fused, and corresponding weights are given to different features to obtain richer feature-maps. The EIoU Loss function is used in the detection layer to optimize the model performance and improve the convergence speed of the prediction frame. Experimental results show that the mean average precision(mAP) of the improved YOLOv5s reaches 91.8%, and the number of frames per second (FPS) is 87.7 f/s, which are 4.6% and 11.4% higher than the original algorithm.
Keywords: slag detection;YOLOv5s;attention mechanism;loss function;BiFPN
2024, 50(7):163-169  收稿日期: 2022-08-01;收到修改稿日期: 2022-09-28
基金项目: 国家自然科学基金资助项目(61763039)；内蒙古自治区自然科学基金项目资助（2022MS06003）
作者简介: 曹君（1995-），男，安徽合肥市人，硕士研究生，专业方向为图像处理、深度学习。
参考文献
"[1] 相伟. 转炉炼钢和电炉炼钢的优缺点比较[J]. 冶金与材料, 2022, 42(2): 45-46.
[2] 饶磊. 转炉钢渣成分、结构及性能间内在规律及其应用研究[D]. 北京: 北京科技大学, 2020.
RAO L. Study on the internal laws of composition,structure and properties of converter steel slag and its application [D]. Beijing: University of Science and Technology Beijing, 2020.
[3] 陈灵光, 杨晓江, 李培玉, 等. 一种改进的转炉出钢下渣检测方法及其应用[J]. 钢铁研究学报, 2012, 24(11): 56-62.
YANG L G, YANG X J, LI P Y, et al. An improved converter tapping slag detection method and its application[J]. Journal of Iron and Steel Research, 2012, 24(11): 56-62.
[4] JULIUS E, THEISSEN W. Function and application of an electromagnetic slag detection system[J]. World Steel Metalwork. Export Man., 1987, 8: 51-54.
[5] 张子淼, 刘斌, 赵盼, 等. 基于红外测温原理的转炉下渣检测系统[J]. 红外与激光工程, 2014, 43(2): 469-473.
ZHANG Z M, LIU B, ZHAO P, et al. Slag detection system based on infrared temperature measurement[J]. Infrared and Laser Engineering, 2014, 43(2): 469-473.
[6] 李爱莲, 刘浩楠, 郭志斌, 等. 改进ResNet101网络下渣出钢状态识别研究[J]. 中国测试, 2020, 46(11): 116-119+125.
LI A L, LIU H N, GUO Z B, et al. Study on the status identification of slag steel under improved ResNet101 network[J]. China Measurement ＆ Test, 2020, 46(11): 116-119+125.
[7] 张幸光. 基于二次阈值分割的红外下渣检测系统研究[D]. 南京: 南京理工大学, 2016.
[8] 何伟. 转炉出钢下渣检测方法与挡渣设备的研究[D]. 哈尔滨: 哈尔滨工业大学, 2017.
HE W. A study on detection m ethod of ladle slag and slag equipment[D].Harbin: Harbin Institute of Technology, 2017.
[9] 谢秋菊, 吴梦茹, 包军, 等. 融合注意力机制的个体猪脸识别[J]. 农业工程学报, 2022, 38(7): 180-188.
XIE Q J, WU M R, BAO J, et al. Individual pig face recognition combined with attention mechanism[J]. Transactions of the Chinese Society of Agricultural Engineering, 2022, 38(7): 180-188.
[10] 陈欣佳, 刘艳霞, 洪晓斌, 等. 基于SSD-CF的无人艇目标检测跟踪方法[J]. 中国测试, 2019, 45(2): 145-150.
CHEN X J, LIU Y X, HONG X B, et al. Unmanned boat target detection and tracking method based on SSD-CF[J]. China Measurement ＆ Test, 2019, 45(2): 145-150.
[11] 彭伟康, 陈爱军, 吴东明, 等. 基于改进Faster R-CNN的水准泡缺陷检测方法[J]. 中国测试, 2021, 47(7): 6-12.
PENG W K, CHEN A J, WU D M, et al. Defect detection method of level bubble based on improved Faster R-CNN[J]. China Measurement ＆ Test, 2021, 47(7): 6-12.
[12] 范玉莹, 何赟泽, 孙高森, 等. 无人机图像风力发电机轮毂中心检测与跟踪[J]. 中国测试, 2022, 48(7): 90-96.
FAN Y Y, HE Y Z, SUN G S, et al. Wind turbine hub center detection and tracking based on UAV images[J]. China Measurement ＆ Test, 2022, 48(7): 90-96.
[13] REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: Unified, real-time object detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016.
[14] REDMON J, FARHADI A. YOLO9000: better, faster, stronger[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017.
[15] REDMON J, FARHADI A. Yolov3: An incremental improvement[EB/OL]. (2018-04-08)[2022-08-01]. https://arxiv.org/pdf/1804.02767.pdf.
[16] BOCHKOVSKIY A, WANG C Y, LIAO H Y M. Yolov4: Optimal speed and accuracy of object detection[EB/OL]. (2020-04-23)[2022-08-01]. https://arxiv.org/abs/2004.10934v1.
[17] LIN T Y, DOLLÁR P, GIRSHICK R, et al. Feature pyramid networks for object detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017.
[18] LIU S, QI L, QIN H, et al. Path aggregation network for instance segmentation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern recognition, 2018.
[19] WOO S, PARK J, LEE J Y, et al. Cbam: Convolutional block attention module[C]//Proceedings of the European Conference on Computer Vision (ECCV), 2018.
[20] ZHOU B, KHOSLA A, LAPEDRIZA A, et al. Learning deep features for discriminative localization[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016.
[21] ZHU X, CHENG D, ZHANG Z, et al. An empirical study of spatial attention mechanisms in deep networks[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision, 2019.
[22] TAN M, PANG R, LE Q V. Efficientdet: Scalable and efficient object detection[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020.
[23] ZHANG Y F, REN W, ZHANG Z, et al. Focal and efficient IOU loss for accurate bounding box regression[J]. arXiv preprint arXiv: 2101.08158, 2021.
[24] LIU W, ANGUELOV D, ERHAN D, et al. Ssd: Single shot multibox detector[C]//European Conference on Computer Vision, 2016.