基于直线特征的铁路场景红外与可见光图像配准方法研究

Visible and Infrared Image Registration in theRailway Environment with Linear Features

列车运行环境状态智能感知是实现铁路列车自主运行控制的基础,而异源图像配准是全天候条件下多源融合感知的前提。本文针对红外与可见光图像多源融合进行铁路全天候环境感知的问题,提出了铁路场景下基于直线特征的红外与可见光图像配准算法。首先利用极向投影法提取图像钢轨直线,然后结合直线特征构建描述符,定义匹配分数公式精确匹配同名直线,实现粗匹配;基于粗匹配产生的候选框提取多方向直线特征进行精匹配;最后根据匹配结果计算模型变换参数完成配准。实验结果表明,该算法计算量小,速度快,配准精度高,为后续的环境感知和目标识别奠定了基础。

Intelligent pereeption of train operating environment is the basis of autonomous train control,and im-age registration is the prerequisite of multi-source fusion perception under all weather conditions.In this paper,an infrared and visible image registration algorithm based on linear features is proposed for railway all-weather en-vironment perception by muli-modal infrared and visible images fusion.First,the straight lines of rails are ex-lracted by the polar projection method,then the descriplor is constructed based on the straight line features.Andthe matching fraction formula is defined to evaluate the homonymous lines to achieve rough matching.A muli-di-rectional straight line feature is extracted based on the candidate box generated by rough matching for precisematching.Finally,the registration is completed by calculating the model transformation parameters based on thematching results.The experiments show that this algorithm is compultationally efficient and has high registrationaccuracy,and provides a good foundation for subsequent environmental perception and objects recognition.