【作者】 冀思远； 贾卓生；

【Author】 JI Si-yuan;JIA Zhuo-sheng;Departmen of Computer and Information Technology,Beijing Jiaotong University;

【机构】 北京交通大学计算机与信息技术学院；

【摘要】 为了解决现有的目标检测方法对5G基站电缆图像检测精度不高的问题,提出了一种多尺度特征融合的目标检测方法(MFN)。MFN用特征融合网络,融合不同深度的特征信息,用多尺度检测网络生成候选区域,以提高检测精度,并在5G基站电缆数据集上证明了 MFN的优越性。与其他方法相比,MFN的均值平均精度以及对地线、火线和光纤3类难区分目标的检测精度均有所提高。更多还原

【Abstract】 In order to solve the problem of low detection accuracy of existing target detection methods for 5G base station cable images,a multi-scale feature fusion target detection method(MFN) is proposed in the paper.MFN uses a feature fusion network,which fuses feature information of different depths,a multi-scale detection network is used to generate candidate regions to improve detection accuracy,and validation on a 5G base station cable dataset shows the superiority of MFN.Compared with other methods,the mean-average accuracy of MFN and the detection accuracy of three types of indistinguishable targets,ground,firewire and fiber,are improved.更多还原