基于钢轨图像频域特征的钢轨波磨检测方法

Detection Method for Rail Corrugation Based on Rail Image Feature in Frequency Domain

通过统计分析钢轨图像中钢轨波磨线和背景线的傅里叶变换系数,发现波磨线的能量集中在频率比较低的区间,而背景线的能量分布比较分散。基于此钢轨图像的频域特征提出了新的钢轨波磨检测方法。首先设计并采用基于位置加权的钢轨定位算法,以快速从轨道图像中提取出钢轨的图像;然后基于钢轨图像每列的傅里叶变换特征,提出频率阈值法和累积能量阈值法2种钢轨波磨线识别算法;最后依据识别出的钢轨波磨线的连续性判定钢轨的波磨区间。结合实际线路的轨道图像,采用新方法和传统方法进行钢轨波磨检测效果的对比试验。结果表明:新方法的精准率和召回率分别为92.19%和97.25%,比传统方法提高了约4%和11%,检测速度也提高了1倍以上。

We statistically analyzed the Fourier transform coefficients of rail corrugation lines and back- ground lines in rail images, and found that the energy of corrugation lines was concentrated in the interval of low frequency but the energy of background lines was often scattered. A new detection method for rail corrugation was proposed based on the rail image features in frequency domain. Firstly, an improved rail location algorithm based on position weight was designed and adopted to extract rapidly rail images from track images. Then, two recognition methods for rail corrugation lines, namely, frequency threshold and accumulated energy threshold methods, were proposed based on the Fourier transform feature of each row in rail image. Finally, the corrugation interval of rail was determined according to the continuity of the recognized rail corrugation lines. Combined with the rail images of actual track, contrast tests were conducted with the proposed method and the traditional one on the detection effect for rail corrugation. Results show that the precision rate and recall rate of the proposed method is 92.19% and 97.25% respectively, which is about 4% and 11% higher than that of the traditional one. Besides, the detection speed has been doubled.