基于bp神经网络的针阔混交林tm遥感图像自动分类技术研究

《基于bp神经网络的针阔混交林tm遥感图像自动分类技术研究》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。

1、Seediscussions,stats,andauthorprofilesforthispublicationat:https://www.researchgate.net/publication/292771651AutomaticallyclassifyingandidentifyingtheTMremotesensingimagesofforestmixedwithconiferandbroadleavesusingimprovedBPANNArticle·January2005CITATIONSREADS782authors,including:ZhengyongZhao

2、UniversitéduQuébecàMontréal22PUBLICATIONS308CITATIONSSEEPROFILEAllcontentfollowingthispagewasuploadedbyZhengyongZhaoon16November2017.Theuserhasrequestedenhancementofthedownloadedfile.第41卷第6期林业科学Vol.41,No.62005年11月SCIENTIASILVAESINICAENov.,2005基于BP神经网络的针阔混交林TM遥感图像自动分类技术研究王立海赵正勇(东北林业大学哈尔滨150040)

3、摘要:在对标准BP神经网络试验分析的基础上,通过输入矢量归一化处理、主成分分析、增加验证集、改进训练学习算法、扩大隐层和输出层规模等措施,对BP神经网络自动分类系统进行改进;利用改进后的BP系统对吉林省汪清林业局的典型针阔混交林TM遥感图像进行辩识、分类试验研究。结果表明:改进后的BP网络分类系统自动分类精度提高了19.14%,比传统无监督自动分类精度提高8.55%,达到了区分森林类型的分类要求。研究还显示了该改进系统应用于针阔混交林TM遥感图像自动分类识别的精度随网络规模增大而提高。关键词:BP神经网络;针阔混交林;TM图像;自动分类;地理信息中图分类号:S771.8文献标识码:

4、A文章编号:1001-7488(2005)06-0094-07AutomaticallyClassifyingandIdentifyingtheTMRemoteSensingImagesofForestMixedwithConiferandBroadleavesUsingImprovedBPANNWangLihaiZhaoZhengyong(NortheastForestryUniversityHarbin150040)Abstract:TheautomaticallyclassifyingandidentifyingtheTMremotesensingimagesofforest

5、playsanimportantroleinthemonitoringandmanagementofforestresources.InordertoimprovetheperformanceofBPartificialneuralnetwork(BPANN),manymeasures,suchasstandardizinginputvectors,increasingverifyingsetvolume,promotingtrainingstudyalgorithm,expandinglayersofinput_outputandmainfactoranalysis,wereap

6、pliedintheTMimagedataprocessing.TakingWangqingForestryBureauofJilinProvinceastheexamplestudyarea,theauthorsstudiedtheautomaticallyclassifyingandidentifyingtheTMremotesensingimagesofforestmixedwithconiferandbroadleavesusingtheimprovedBPANN.Theresultsshowthataccuracyofautomaticallyclassification

7、andidentificationhasbeenincreasedsignificantly,19.14%higherthanthatofthetraditionalANNmethodand8.55%higherthanthatoftraditionalunsupervisedclassifyingmethodrespectively.Theresearchalsoindicatesthattheclassificationandidentificationaccur