基于卷积神经网络的水稻叶片病害检测与识别研究进展

打开文本图片集

中图分类号：S435.1；TP391.4 文献标识码：A 文章编号：2095-5553（2025）10-0176-08

Abstract：The preventionandcontrolof rice leaf diseases inour countryhas always beenatop priority.Therealizationof fastandaccurate disease detectionand classificationidentificationcanhelp todetectandtaketreatment measuresatanearly stage inatimely manner，thus improving theyieldand qualityofrice.Byanalyzing the existing riceleaf disease detection andrecognitionalgorithms，itisfound thatleafdiseasedetectionbasedontraditional imageprocesing methodsisineficient andinaccurate.However，with thecontinuous developmentof deep learning，disease detectionand recognition basedon convolutional neuralnetwork has becomeanimportant topic forresearchers.Aimingatthe model algorithms usedinrecent years，this paper summarizes the improvement strategies such as data preprocessing and data enhancement，framework structure improvement and transfer learning，compares andanalyzes the performance and limitations of these algorithms， andfinds that most models have the problem of imbalance between acuracy and model parameter quantity performance. Finall，the futureresearch trendsare prospected from the aspectsofdatasetconstruction，modelperformance balanceand generalizationability，whichcanprovidereferenceforeficientdetectionand identificationofriceleaf diseases inthe future.

Keywords：riceleaf;diseasedetectionandidentication；convolutionalneuralnetwork；targetdetection;classicationand identification；improvement strategy

0 引言

水稻作为世界上重要的粮食作物之一，不仅是我国，也是许多其他国家重要的口粮[1]。（剩余15641字）