一种针对路面损坏识别的点云特征图像生成方法

A Point Cloud Feature Image Generation Method for Pavement Damage Recognition

以车载激光点云数据为研究对象,提出一种针对路面损坏识别的点云特征图像生成方法。该方法首先对路面点云统计去噪,然后根据路面损坏的空间分布特征,设计一种损坏特征参考插值算子,构建圆形结构高程梯度差分窗口进行路面梯度分析,最后运用中值滤波算子滤除梯度椒盐噪声,生成路面损坏特征图像。经实验分析,对于不同类型的路面点云,该方法生成的路面损坏特征图像可以有效表达路面损坏目标,与传统方法相比,本文算法的提取准确率和召回率分别达到88.89%和92.31%,具有较强的适用性和可靠性。

Taking the laser point cloud data of vehicle as the research object,a method of generating point cloud feature image for road damage identification is proposed.In this method,firstly,the point clouds of pavement are statistically denoised,then a damage feature reference interpolation operator is designed according to the spatial distribution characteristics of pavement damage,and a circular structure elevation gradient difference window is constructed to analyze the pavement gradient.Finally,the median filter operator is used to filter the salt-pepper noise and generate the pavement damage feature image.Experiments show that,for different types of pavement point clouds,the road damage feature images generated by this method can effectively express the road damage targets.Compared with the traditional methods,the extraction accuracy and recall rate of this algorithm are 88.89% and 92.31%,respectively.It has strong applicability and reliability.