自然环境下基于颜色聚类和颜色距离的死钩检测

Fixed-hook detection based on color cluster and color distance in natural environment

针对目前火车死钩检测无法自动实现的问题,提出了一种自然环境下基于颜色聚类和颜色距离的死钩检测方法。根据死钩和车厢颜色的对应关系,使用CCD(charge-coupled device)相机获取现场车厢图像并提取前景区域和背景区域的颜色特征,通过分析该颜色信息的差异来判断车厢之间的连接是否为死钩。首先获取特定区域的颜色信息,然后采用FCM(fuzzy C-mean)聚类算法对颜色信息进行分类得到该区域的单一颜色特征,最后根据HLC(hue,lightness,hromatic)颜色空间和人类颜色视觉的相似关系,计算颜色特征对的NBS(national bureau of standards)颜色距离。利用翻车作业现场火车车厢图像进行检测,实验结果验证了该方法具有对颜色差异的高敏感性和识别的准确性,可以满足实际死钩检测的需要。

In order to solve the problem that train fixed-hook detection cannot be realized automatically,a fixed-hook detection method based on color cluster and color distance is proposed in natural environment.According to the corresponding relationship of fixed-hook and carriage color,a CCD（charge-coupled device） camera is used to acquire the images of the carriages,and the color features of foreground region and background region are extracted.Through analyzing the difference between the color information,it is determined that the connection between the carriages is fixed-hook or not.First,the color information of particular area is acquired;and then the regional single color characteristic is obtained using FCM（fuzzy C-mean） clustering algorithm to classify color information;at last,the NBS（national bureau of standards） distance is calculated according to the similar relationship of color characteristics between Munsell-HLC color space and human color vision.The pictures taken in a train overturn site were used for experiment test;experimental results demonstrate that this method has extremely high sensitivity to color difference and accuracy of color recognition,which can meet the requirement of actual fixed hook test.