摘要
根据目前点云数据分割的研究现状以及分水岭算法在图像处理中的应用,提出一种基于深度图像和分水岭算法的建筑物平面点云分割方法。该方法首先将平面点云数据生成深度图像并给像素赋予灰度值,再使用双边滤波算法进行滤波去噪,然后使用分水岭算法进行图像分割,得到分割结果后索引回原始点云数据,得到点云分割结果。为验证方法的可靠性与准确性,利用区域增长法、RANSAC算法以及欧式聚类法进行对比实验。通过对实验结果的对比分析,能有效地将不同点云面片分割出来,并且具有良好的准确度和完整度,分割结果质量较高,为点云的分割提供了新的思路与方法。
According to the current research status of point cloud data segmentation and the application of watershed algorithm in image processing,this paper proposes a method of building plane point cloud segmentation based on depth image and watershed algorithm. Firstly,the depth image is generated from the plane point cloud data,and the gray value is given to the pixels. Then,the bilateral filtering algorithm is used for filtering and denoising. Then,the watershed algorithm is used to segment the image. After the segmentation results are obtained,the original point cloud data are indexed back to get the point cloud segmentation results. In order to verify the reliability and accuracy of the method,the regional growth method,RANSAC algorithm and Euclidean clustering method are used for comparative experiments. Through the comparative analysis of the experimental results,this method can effectively segment different point cloud patches,and has good accuracy and integrity,and the segmentation result quality is high,which provides a new idea and method for point cloud segmentation.
作者
刘瑞
吕开云
袁志聪
王凯
LIU Rui;LYU Kaiyun;YUAN Zhicong;WANG Kai(Faculty of Geomatics,East China University of Technology,330013,Nanchang,PRC;Zhuhai Surveying and Mapping Institute,519000,Zhuhai,Guangdong,PRC)
出处
《江西科学》
2021年第1期166-171,共6页
Jiangxi Science
基金
国家自然科学基金(42061077,42064001)
国家重点研发计划(2016YFB0501405,2016YFB0502601-04)
江西省自然科学基金(2017BAB203032)。
关键词
深度图像
双边滤波
分水岭算法
图像分割
点云分割
depth image
bilateral filtering
watershed algorithm
image segmentation
point cloud segmentation
