基于卷积网络的受电弓图像目标检测与矫正方法研究

Study on Pantograph Image Target Inspection and Correction Method based on Convolution Network

受电弓滑板监测装置(5C)存在安装角度偏差,其拍摄的受电弓图像易出现透视畸变,难以进行后续检测。本文提出一种受电弓检测与图像矫正方法,首先利用残差网络(ResNet)提取图像特征,再通过转置卷积网络提升特征图分辨率,最后利用定位得到受电弓4个顶点的坐标,完成受电弓的定位和透视矫正。实验分析了不同残差网络结构以及不同输入图像尺寸对受电弓顶点检测准确率的影响,结果表明,本文所述方法能够有效定位受电弓,且图像矫正效果良好。

The monitoring device for contact stripe of pantograph(5 C)has an installation angle deviation,and the pantograph image taken by 5 C will appear perspective distortion,making it difficult to carry out subsequent inspections.Therefore,the paper proposes a method for pantograph positioning and image correction.Firstly,the residual network(ResNet)is used to extract the features of the image,and then the transposed convolutional network is used to improve the resolution of the feature map.Finally,the coordinates of the four vertices of the pantograph obtained by positioning are used to complete the positioning and perspective correction of the pantograph.The experiment analyzes the influence of different residual network structure and different input image size on the accuracy of pantograph vertex inspection.The experimental results show that this method can effectively position the pantograph,and the images are well corrected.