基于DBSCAN-FPN的输电线路螺栓缺销检测方法

螺栓作为输电线路上数量最大的紧固件，其缺陷检测是输电线路巡检工作中的一项重要内容。针对螺栓缺销为小目标，其定位困难、特征难提取的问题，提出一种基于DBSCAN算法与FPN模型相结合的螺栓缺销检测方法。首先，利用FPN模型定位螺栓缺销目标区域，同时基于DBSCAN聚类算法对具有相同形态结构的区域进行聚类；然后，改进FPN模型：基于螺栓先验知识，利用卷积网络实现自底向上的特征提取，采用双线性插值方法将特征的高层语义信息自顶向下地传递到各个层级，通过卷积滤波方法横向加强高层语义特征与高分辨率特征的融合信息，获得更优化的螺栓缺销特征表达；利用改进FPN模型实现螺栓缺销的初步检测；最后，采用DBSCAN聚类算法对初步检测结果进行误检甄别，实现了螺栓缺销的精确检测。实验结果表明，DBSCAN-FPN在自建数据集上的检测精度达到76.23%，检测效果优于FPN、R-FCN和Faster R-CNN。所提方法可以有效提高螺栓缺销检测精度，对输电线路运维有实际意义。

Detection Method for Bolts with Mission Pins on Transmission Lines Based on DBSCAN-FPN

Bolts are the mostly used fasteners on transmission lines,and their defect detection is an important content for transmission line inspection.As the bolt with missing pins are small targets,their positioning is difficult and their features are hard to extract.Aim at this problem,a detection method for bolts with missing pins is proposed based on the DBSCAN algorithm and FPN model.Firstly,the FPN model is used to locate the target area of the bolts with missing pins,and the areas with same morphological structure are clustered based on the DBSCAN clustering algorithm.Then,the FPN model is improved:based on the prior knowledge of bolts,the convolutional network is used to achieve bottom-up feature extraction,the bilinear interpolation method is used to transfer the high-level semantic information of the features to each level from top to bottom,the convolution filtering method is used to laterally strengthen the information fusion of high-level semantic features and high-resolution features,and a more optimized feature expression of bolts with missing pins is obtained.The improved FPN model is used to realize the preliminary detection of the bolts with missing pins.Finally,the DBSCAN clustering algorithm is adopted to screen the preliminary detection results for error detection,thereby achieving the accurate detection of the bolts with missing pins.Experimental results show that the detection accuracy of DBSCAN-FPN reaches up to 76.23%on the self-built data set,with the detection effect better than FPN,R-FCN and Faster R-CNN.The proposed method can effectively improve the detection accuracy of bolts with missing pins,which has practical significance for the operation and maintenance of transmission lines.