基于视频图像的弓网接触位置动态监测方法

Dynamic Monitoring Method of Pantograph-Caternary Contact Position Based on Video Image

针对列车人工检测受电弓状态效率低下的问题,提出基于视频图像的弓网接触位置动态监测方法。首先,将运用帧间差抓取受电弓视频中得到的包含弓网的图像作为原始训练数据集;然后,对复杂的受电弓图像背景进行分割处理,采用超像素分割获得最大特征区域,结合特征图像的HOG(方向梯度直方图)获得最大特征ROI(感兴趣区域),形成训练数据集,并设计标签;最后,运用改进的YOLO v3-tiny-strong网络结构检测分类器,用训练的权重对视频目标进行监测。结果表明:该动态监测方法能够在每1帧图像中精确标记出受电弓与接触网的接触点位置,并且能够持续对受电弓的运动状态进行捕捉,有效获取接触点与受电弓的相对坐标位置,从而达到对受电弓的监测目的,其对弓网视频的检测精度可达98%。

Aiming at the problem of low efficiency of train manual detection of pantograph status,a dynamic monitoring method of pantograph-catenary contact position based on video image is proposed.First,the image containing the pantograph-catenary obtained from the pantograph video captured by frame difference is used as the original training data set;then,the complex pantograph image background is segmented,and the super pixel segmentation is used to obtain the largest feature area,combining the HOG(Histogram of Directional Gradients)of the feature image to obtain the largest feature ROI(region of interest),to form the training data set,and to design the label;finally,the improved YOLOv3-tiny-strong network structure is used to detect the classifier and trained weight is used to monitor video targets.Results show that the dynamic monitoring method can accurately mark the position of the contact point of pantograph and catenary in each frame of image and can continuously capture the movement state of the pantograph,and effectively obtain the contact point and the pantograph relative coordinate position,so as to achieve the purpose of monitoring the pantograph,of which the detection accuracy can reach 98%.