基于轻量化YOLOv8-Rice的库尔勒香梨虫害监测研究

打开文本图片集

Abstract： Aimingatthechallngesofcomplexckgrond，largecalulatioamountandlageparametersoftheoeladdifult tobedeployedonmobileorembeddeddevicesinthientificationofpestsanddiseasesofKorlaFragrantPear，thistudyproposeda improvedlightweight model （YOLOv8-Rice）basedonYOLOv8.Frstly，this modelusedtheContext GuidedBlockstructure to replacetheBotlecheckstructureoftheC2fmoduleinYOLOv8，whichefectivelyenhancedtheunderstandingabilityofthemodel to the context informationandcompreses the model weight.Then，thestandard convolution in YOLOv8 was replaced bythe depth-separatableconvolution，whichsignfcantlyreducedtheparametersandcalculationofthemodelFinally，byreconfigurng thedetectionheadasalightweightsharedconvolutiondetectionhead，the parameterandcalculationamountofthe model were furtherreducedandepositiongdtraconcpaityftemodelfortheactesticsultialeseasessts was improvedTheexperimentalresults show that，compared withYOLOv8，thecomputationaland parameterquantities of YOLOv8-Rice are reduced by 71.1% and 61.5% respectively，and that the model weight file size is reduced to 1.89 MB， only 32.4% of YOLOv8. Meanwhile，the average accuracy reaches 94.0% which issignificantly improved compared with other models.

Key Words： Pest and disease detection; Deep learning; Korla Fragrant Pear; Lightweight

doi：10.13620/j.cnki.issn1007-7782.2025.02.008

中图分类号：S24;TP2 文献标识码：A

0引言

库尔勒香梨作为新疆地区特色水果，对当地农业经济具有显著影响。（剩余6179字）