基于SVM算法的植被DIM点云提取方法研究

打开文本图片集

中图分类号：P23 文献标识码：A文章编号：1006-8228（2025）12-31-06

Research on DIM Point Cloud Extraction Method of Vegetation Based on SVM Algorithm

CuiYing，Li Chunmei

（Shanxi RailwayVocational and Technical College，Taiyuan，Shanxi O3oooo，China）

Abstract：Thestudydrawsontheideaofsamplelearninginmachinelearning，andproposedavegetationdenseimagematching （DIM）pointcloudextractionmethodbasedonthesupportvectormachine（SVM）algorithm.FirstlyasmallnumberofDMpoint cloudsitheexperimentalareawereselectedasthesampleset，andthevegetationpointcloudsandnon-vegetationpointclouds withinthesamplesetweremarkedastwocategories.Then，featurevectorswereconstructedbasedonthecolorinformationofthe samplepoint clouds，andtheSVMalgorithmwasusedtoconstructtheclassficationmodeltolearnthesamplefeatures.Model trainingassuccesfullycompletedbyiterativecalculationtodeterminethesupportvectorsandsegmentationhyperplanerquired forthemodel.Experimentalresultsshowthatthemodelcanachieveeficientvegetationpointcloudextractionindiferentscenes. Among them，the shortest convergence time of the model in the school scene can reach 0.9s ，andthenumberof training iterations isonly 4 times，with a recognition accuracy ofover 94% .Therefore，the classification model based on the SVM algorithm for extractingvegetationDIMpointcloudshasstableconvergencespeedandrecognitionaccuracy，whichcanprovidetechnical reference for the application research on machine learningin point cloud classfication and extraction.

eyWords：Machine Learning;SVMAlgorithm;Sample Features;Vegetation DIMPoint Cloud

0引言

人工智能时代的到来，促使测绘科学从几何科学逐步向信息科学发展，对地理空间信息数据的分析和挖掘提出了新挑战-3]。（剩余7568字）