基于激光雷达的无人机导航图像分水岭快速分割方法  

A fast watershed segmentation method for UAV navigation image based on

作　　者：杨治 彭蕾[1] 涂起龙[1] LiDAR YANG Zhi;PENG Lei;TU Qilong(School of Electronic Information Engineering Jinggangshan University,Ji’an Jiangxi 343009,China)

机构地区：[1]井冈山大学电子信息工程学院,江西吉安343009

出　　处：《激光杂志》2023年第8期125-129,共5页Laser Journal

基　　金：教育厅科学技术研究一般项目(No.GJJ211031)。

摘　　要：为了更准确地划分地物、分类图像中的内容,提出基于激光雷达的无人机导航图像分水岭快速分割方法。首先,通过高斯滤波的方式平滑处理原始图像,再利用形态学梯度算子计算其梯度,消除梯度图像中的明暗干扰;然后,将灰度相似度与纹理对比度作为衡量指标,搜索像素内灰度级的排序特征以及纹理在图像目标的转变;最后,应用激光雷达点云数据的高程信息优化分割,检测目标没有关联的最小值点,明确处理区的相邻关系,获得精准的分割结果。实验表明,该方法能够准确、快速地完成图像分割,分割后的图像整体布局均匀、轮廓清晰、分类清晰。In order to more accurately divide the ground objects and classify the contents of images,a fast water-shed segmentation method based on LiDAR for UAV navigation images is proposed.First,the original image is smoothed by Gaussian filtering,and then its gradient is calculated by morphological gradient operator to eliminate the light and dark interference in the gradient image;Then,the gray level similarity and texture contrast are used as meas-urement indexes to search for the sorting features of gray levels in pixels and the transformation of texture in image ob-jects;Finally,the elevation information of the LiDAR point cloud data is used to optimize segmentation,detect the minimum points that are not associated with the target,clarify the adjacent relationship of the processing area,and ob-tain accurate segmentation results.Experimental results show that this method can complete image segmentation accurately and quickly,and the segmented image has uniform overall layout,clear outline and clear classification.

关 键 词：激光雷达 无人机导航 分水岭算法 图像分割 梯度图像 

分 类 号：TN958.98[电子电信—信号与信息处理]