摘要
在对接触网悬挂装置状态进行检测时,因为缺少开口销缺失的样本,智能识别难以保证较高的准确率。本文提出了一种基于YOLO v2检测算法与深度降噪自编码网络的接触网开口销缺失识别方法,实现对开口销的定位及定位后开口销图像重构,通过分析重构误差判断接触网开口销是否缺失。试验证明该方法能有效识别开口销缺失情况。
Due to lack of samples of missing split pins, it is difficult to guarantee higher accuracy rate for intellectual identification made during state inspection for OCS suspension devices. The paper puts forward identification method for OCS split pins on the basis of YOLO v2 inspection algorithm and deep de-nosing self-coding network to realize the positioning of split pins and reconfiguration of split pins after positioning, the analysis of reconfiguration error is used to confirm whether the OCS split pins are missing or not. The test verifies that the method is able to identify the missing of split pins.
出处
《电气化铁道》
2019年第3期43-47,共5页
Electric Railway
基金
国家重点研发计划(2017YFB1201202)
