针对高速铁路接触网支撑结构中旋转双耳耳片断裂故障难以检测的问题,提出一种HOG(histogram of oriented gradients,梯度方向直方图)特征与二维Gabor小波相结合的图像检测方法。为实现旋转双耳在待检测图像中的定位,利用其正负样本的HOG特征对线性SVM分类器进行训练,对检测窗口内是否包含旋转双耳进行判别。为实现耳片断裂故障的可靠诊断,利用二维Gabor小波变换能量值对图像中的边缘信息进行筛选,进而对耳片断裂故障引起的故障裂痕进行识别。实验结果表明,本文提出的方法能在复杂的接触网支撑与悬挂装置图像中准确识别发生耳片断裂故障的旋转双耳部件,检测结果不受拍摄距离、拍摄角度以及曝光度等因素的影响,具有较高的使用价值。
This paper proposed an image method to detect the fracture failure of the ear pieces of the cross link clevises in the support structure of high-speed railway catenary based on histogram of oriented gradient(HOG)features combined with 2D Gabor wavelet transform.For the extraction of clevises,a linear SVM classifier was trained based on the HOG features of both the positive and negative samples to identify the clevis in the detection window.For the reliable diagnosis of the fracture failure of ear pieces,the energy value of 2D Gabor wavelet transform was used to sift the edge information extracted from the subimage,and to further detect the crack of fracture failure.Experiment results showed that the proposed method has a high utility value as it can realize the accurate identification of clevises with a fracture failure of the ear pieces,not affected by the shooting distance,shooting angle and the illumination intensity.