基于改进型SSD算法的钢缆表面缺陷检测方法  

Surface defect detection method of steel cable based on improved SSD algorithm

作　　者：姜冲 莫文洁 冯上榜 杨永杰[1,2] 许鹏 JIANG Chong;MO Wenjie;FENG Shangbang;YANG Yongjie;XU Peng(School of Information Science and Technology,Nantong University,Nantong 226019,China;Nantong Research Institute for Advanced Commuication Technologies Co.,Ltd.,Nantong 226019,China)

机构地区：[1]南通大学信息科学技术学院,江苏南通226019 [2]南通先进通信技术研究院有限公司,江苏南通226019

出　　处：《电子设计工程》2025年第7期66-71,共6页Electronic Design Engineering

基　　金：国家自然科学基金项目(62271271);江苏省重点研究发展计划项目(BE2021013-1)。

摘　　要：针对在框绞过程中钢缆表面会产生缺陷的问题,采用了改进型SSD算法的方法。具体措施为:使用ResNet18作为SSD的骨干算法来提取特征信息;通过萤火虫优化算法和K-Means算法优化先验框,提高匹配精度;引入单向特征融合模块和改进的CBAM注意力模块,提高了检测的精度;将Focalloss作为损失函数,减少训练过程中负样本的权值。结合钢缆缺陷数据集进行训练实验,得出改进型SSD算法平均精度为80.3%,相对于传统的SSD-VGG模型提升9%,FPS保持在63.3。在检测精度和检测速度方面上,改进型SSD算法能满足实际需求。In order to solve the problem of defects on the surface of steel cable during the production of frame twisting,an improved SSD algorithm is adopted.The specific measures are as follows:ResNet18 is used as the backbone algorithm of SSD to extract the feature information;Firefly algorithm and KMeans algorithm were used to optimize the prior frame and improve the matching accuracy;The unidirectional feature fusion module and improved CBAM attention module are introduced to improve the detection accuracy;Focalloss is used as a loss function to reduce the weight of negative samples during training.The training experiment was carried out by using the defect data from steel cable.It is concluded that the average accuracy of the improved SSD algorithm is 80.3%,which is 9%higher than that of the SSD-VGG model,and the FPS is maintained at 63.3.In terms of detection accuracy and detection speed,the improved SSD algorithm can meet the actual needs.

关 键 词：钢缆缺陷 注意力机制 目标检测 SSD算法 特征融合 

分 类 号：TN2[电子电信—物理电子学]