改进型C4.5算法在多波束点云自动消噪中的应用  

Application of improved C4.5 algorithm to automatic noise reduction in multi-beam point cloud

作　　者：史俊明 卢清国 SHI Junming;LU Qingguo(Department of Spatial Information Engineering of Henan College of Surveying and Mapping,Henan,Zhengzhou,451464,China;Map Institute of Henan Province,Henan,Zhengzhou,450003,China)

机构地区：[1]河南测绘职业学院空间信息工程系,河南郑州451464 [2]河南省地图院,河南郑州450003

出　　处：《海洋测绘》2025年第1期21-24,30,共5页Hydrographic Surveying and Charting

摘　　要：传统C4.5算法广泛应用于多波束测深系统中水深点与噪声点的自动分离,但是该算法中存在过拟合和属性相关性问题。为解决传统C4.5算法的缺陷,引入改进型C4.5算法,提出了通过计算属性依赖度更新数据集合的方法。首先通过点云之间的点特征、区域特征、窗口特征等建立数据集合,并对连续特征属性进行离散化处理。然后构建决策树,利用信息增益率判断根节点(即确定最优属性划分),迭代更新,完成数据集合的分类,并对整体数据进行后剪枝处理。采用多种地形对本文算法训练效果进行交叉评定,证明本文算法的泛化能力和准确度较高。The traditional C4.5 algorithm is widely used in automatic separation of water depth and noise points in Multibeam Bathymetry System,but there are problems of overfitting and attribute correlation in the algorithm.In order to solve the defects of traditional C4.5 algorithm,the improved C4.5 algorithm is introduced in this paper,it presents a method to calculate the attribute dependency update data set.Firstly,the data set is established through the point features,regional features and window features among the point clouds,and the continuous feature attributes need to be discretized.Then the decision tree is constructed,the root node is judged by the information gain rate(that is,the optimal attribute division is determined),the data set is iteratively updated,the classification is completed,and the whole data is post-pruned.The training effect of the proposed algorithm is verified by various terrain,which proves that the proposed algorithm has high generalization ability and accuracy.

关 键 词：海洋测绘 多波束测深系统 改进型C4.5算法 信息增益率 属性依赖度 后剪枝 

分 类 号：P229[天文地球—大地测量学与测量工程]