基于密集网络的中草药识别系统设计与实现

打开文本图片集

中图分类号：TP317；TP183 文献标识码：A 文章编号：1006-8228（2025）12-77-05

Abstract：TraditionalChineseherbsembodyrichculturalheritageandpharmacologicalvalue.Withtherapidadvancementof modernmedicineandartfcialintellgencetechnologies，implementingautomatedidentificationoftraditionalChineseherbsusing deeplearningmethodsrepresentsacrucialresearchdirectionforthedigitalizationandinteligenttransformationoftraditional medicine.Addressingchallngessuchasthevastvarietyofherbs，theirsimilarappearances，andloweficiencyinmanual identification，thispaperproposesatraditionalChinesehrbrecognitionsystembasedontheDenseNet121model.Developedusing PythonandtheFlaskframework，thesystemenablesimageuploading，displayofrecognitionresults，andherbalinformationqueries. ItconstructsadatasetbycrawlingandcolletingactualherbalimagesFollowingimagepreprocessng，theDenseNetmodel performsfeatureextractioclasificatioandecogitionExperimentsemonstrateattheenselycoectedarchitectrefte modelcaneectivelymitigatethegradientanishingproblem，improvefeaturereuseficiencyandrecognitionaccuracywiththe test set recognition accuracy reaching 94.44% ：

Keywords：ChineseHerbalMedicineIdentification;DeepLearing;DenseNet121;ImageClassification;ArtificialInteligence

0引言

中草药种类繁多、形态相似，传统识别方法依赖专业人员的知识和经验，不仅主观性强，且识别效率低下，易出现误识别问题。（剩余6630字）