基于特征融合的接触网定位支座区域检测研究

Research on Contact Area Detection of Contact Network Based on Feature Fusion

针对电气化铁路接触网检测领域中,定位支座区域快速准确检测的问题,分别使用SURF加速稳健特征匹配的方法、局部二值模式LBP特征结合支持向量机SVM的方法和方向梯度直方图HOG特征结合SVM的方法对接触网定位支座区域进行对比试验。在此基础上,提出将LBP特征和HOG特征降维处理后,串行融合结合SVM的方法对定位支座区域进行检测。试验结果表明,在同等条件下,采用此方法的检测准确率和检测效率优于单独使用SURF特征匹配、LBP特征和HOG特征进行检测的方法,能够较好的满足电气化铁路接触网检测的需求。

In the field of electrified railway contact network detection,the problem of rapid and accurate detection of the positioning of the bearing area is to use the SURF accelerated robust feature matching method,the local binary mode LBP feature combined with the support vector machine SVM method and the direction gradient histogram HOG feature combination.The SVM method compares the contact area of the contact net.On this basis,it is proposed that the LBP feature and the HOG feature are reduced in dimension,and the serial fusion com⁃bined with the SVM is used to detect the positioning of the bearing area.The experimental results show that under the same conditions,the detection accuracy and detection efficiency of this method are better than those of SURF feature matching,LBP feature and HOG feature,which can better meet the needs of electrified railway contact network detection.