考虑时间序列模糊分割的多波段激光数据分类  

Multi-band laser data classification considering fuzzy segmentation of time series

作　　者：刘艳 黄亚博[2] LIU Yan;HUANG Ya-bo(College of Information Engineering,Zhengzhou University of Technology,Zhengzhou 450044,China;College of Computer and Information Engineering,Henan University,Kaifeng 475004,China)

机构地区：[1]郑州工程技术学院信息工程学院,河南郑州450044 [2]河南大学计算机与信息工程学院,河南开封475004

出　　处：《激光与红外》2025年第4期520-525,共6页Laser & Infrared

基　　金：河南省高等教育教学改革研究与实践项目(No.2024SJGLX0570);河南省科技攻关计划(工业领域)项目(No.212102210571)资助。

摘　　要：激光扫描仪在不同波段上的测量受到大气条件、目标表面等多种因素的反射特性影响,使得数据在波段间存在交又敏感性,导致数据在时间序列上表现出不确定性,从而降低了激光数据的分类精度。为此,提出考虑时间序列模糊分割的多波段激光数据分类。联合主成分分析方法以及遗传算法,提取多波段激光数据的特征,并找出其中最大特征量,构建多波段激光数据的时间序列;使用遗传算法对激光数据序列实施优化处理,减少了交叉敏感性对分类结果的影响;使用模糊分割算法将优化后的激光数据序列分割成若干时间序列段,联合K-means算法完成序列段聚类,实现多波段激光数据的精准分类,增加了多波段激光数据的分类精度。实验结果表明,利用该方法开展多波段激光数据分类时,分类精度高、分类效果好。The measurement of laser scanner in different bands is affected by the reflection characteristics of atmos-pheric conditions,target surface and other factors,which makes the data cross-sensitivity between bands,resulting in data exhibiting uncertainty in the time series,thus diminishing the classification accuracy of the laser data.For this reason,a multi-band laser data classification considering time series fuzzy segmentation is proposed.Firstly,combined with principal component analysis and genetic algorithm,the features of multi-band laser data are extracted,the maxi-mum feature vector is found out,and the time series of multi-band laser data is constructed.Secondly,the genetic algo-rithm is employed to optimize the laser data sequence,which mitigates the impact of of cross sensitivity on the classifi-cation results.Then,the fuzzy segmentation algorithm is utilized to divide the optimized laser data sequence into sever-al time series segments,and in conjunction with the K-means algorithm,sequence segment clustering is completed to achieve precise classification of multi-band laser data and enhance its classification accuracy.The experimental results show that this method yields high classification accuracy and excellent classification effects when applied to multi-band laser data classification.

关 键 词：时间序列 模糊分割 多波段激光数据 K-MEANS算法 主成分分析方法 

分 类 号：TN914[电子电信—通信与信息系统] TP391[电子电信—信息与通信工程]