一种改进的铁路扣件图像校正算法

基于HOG特征的铁路扣件检测算法中,各类扣件特征的可辨性易受光照不均匀因素影响,降低了检测准确率.针对此类光照不均匀图像,将二维经验模态分解(BEMD)理论与直方图均衡化相结合,首先利用BEMD变换提取图像的细节分量和照射分量,并通过直方图均衡化增强细节分量;然后进行光照判断,据光照判断对细节增强分量叠加调整后的照射分量,得到光照均匀、细节增强的扣件图像.真实实验表明,与传统方法相比,此方法能有效增强图像对比度,提高铁路扣件检测精度.

An improved correction algorithm for railway fasteners image

Railway fastener recognition algorithm based on Hierarchical Histograms of Oriented Gradients（HOG） was affected by the illumination factor easily. For this case, this paper presented an algorithm which was combined the bidimensional empirical mode decomposition（BEMD） with histogram equalization. Firstly, the illumination component and detail component were extracted by BEMD. Detail component was enhanced by histogram equalization, and illumination component was revised based on illumination judgment, this algorithm would get the satisfactory fastener image for the further detection. Compared to the traditional algorithm, experimental results showed that this algorithm could improve image contrast and recognition accuracy.