基于CBAM的多模光纤微小位移状态检测方法  

Detection Method for Micro Displacement State of Multimode Fiber Based on CBAM

作　　者：王京 翁梦缘 王真真 WANG Jing;WENG Mengyuan;WANG Zhenzhen(Hebei University of Architecture,Zhangjiakou 075000,China)

机构地区：[1]河北建筑工程学院,河北张家口075000

出　　处：《仪表技术》2024年第4期55-59,共5页Instrumentation Technology

摘　　要：传感技术在传统行业领域已经得到了广泛应用,但受外部干扰影响,在微小目标检测和高精度需求方面略显不足。为了实现多模光纤微小位移状态的检测,提出了一种基于YOLOv4-tiny-CBAM模型结构的光斑识别方法。利用光纤弯曲程度与输出光斑之间的对应关系,通过采集多模光纤不同位移条件下的光斑图像,并结合卷积块注意力模块(CBAM)进行深度学习网络训练。实验结果表明,该方法在不进行预训练的情况下,将模型的平均精度从原模型的42%提高到了82.95%,实现了对多模光纤微小位移状态的高效识别。Sensing technology has been widely applied in traditional industries,but due to external interference,it is slightly insufficient in terms of small target detection and high-precision requirements.A spot recognition method based on the YOLOv4 tiny CBAM model structure is proposed to achieve the detection of small displacement states in multimo-de optical fibers.The correspondence between the degree of fiber bending and the output spot is used to collect spot images of multi-mode optical fibers under different displacement conditions,and combine with the convolutional block attention module(CBAM)for deep learning network training.The experimental results show that this method improves the average accuracy of the model from 42%of the original model to 82.95%without pre-training,achieving efficient identification of small displacement states of multimode optical fibers.

关 键 词：注意力机制 卷积块注意力模块 深度学习 光纤光斑 位移 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]