基于视频图像的轨道前方障碍物检测方法

Track image obstacle detection method based on video image

提出了一种基于视频图像的列车前方障碍物检测方法。首先对拍摄的图像进行图像数据预处理;其次,根据二值化图像进行轨道拟合检测并建立检测图像窗口;然后对预处理后的图像进行反透视变换处理,通过分析反透视变换处理后的图像对其像素值做列向累加;根据峰值划分横向区域,对分区图象的像素值做横向累加,综合确定障碍物占据的参数;最后,进行逆反透视变换标记出障碍物具体的位置。实验结果表明,本方法对于直道场景下的铁路障碍物检测能有效地检测出障碍占据,并判断出障碍类型,可以有效降低人工检测的困难和危险,并提高效率。

A method based on video image for obstacle detection in front of train is proposed. Firstly,image data is preprocessed for the captured image. Secondly,the orbital fitting detection is performed according to the binarized image and the detection image window is established. Then perform inverse perspective transformation on the preprocessed image,by analyzing the inverse-transformed image,the pixel values are column-accumulated;the horizontal region is divided according to the peak,and the pixel values of the partition image are horizontally accumulated. and the parameters occupied by the obstacle are comprehensively determined;finally,A reverse-reverse perspective transformation is performed to mark the specific location of the obstacle. The experimental results show that the method can effectively detect the obstacle occupation and determine the obstacle type for the detection of railway obstacles in the straight scene,which can effectively reduce the difficulty and danger of manual detection and improve the efficiency.