高光谱技术检测土壤养分研究进展

打开文本图片集

中图分类号：S224.2 文献标识码：A 文章编号：2095-5553（2026）04-0214-08

Abstract：Asanimportantpartofprecisionagriculture，precisionfertilizationisgraduallbeingpromotedandapplied in China'sagriculturalfield.Monitoringsoilnutrientcontentisthefirststepinprecisefertilization，andhyperspectralimaging technologyisconsideredas the most idealmethod，whichcan identifyand quantify key nutrients such as organic mater， nitrogen，phosphorus，andpotassum inthesoil.This paperreviews thelatestresearch progressofhyperspectral imagingin thefieldof soilnutrient monitoring，anddeplydiscusses diferent spectral preprocessingalgorithms，characteristicband selection，theconstructionofregressonprediction models，and theinfluenceofsoil physicalandchemicalparameterson the monitoringresults.Throughcomparativeanalysis，thispaper finds thattherearestillsomekeyproblems to be solved urgentlyin termsof thedata processing eficiency，adaptabilityof prediction modelsand monitoring accuracyof existing algorithms，suchas noiseremoval of spectral signals，redundancyofcharacteristic bands，and insuficientgeneralization abilityof models.Basedonthis，itisproposed that in thefuture，the focusshould beonmoreefcient spectral preprocessing algorithms，accurate feature band selectionalgorithms，and regresion prediction model optimizationcombined withdep learning，etc.，aiming topromotetheapplicationandpopularizationof hyperspectral imaging technology in agriculturalfine management and provide more reliable technical support for accurate monitoring ofsoil nutrients.

Keywords： soil nutrients; hyperspectral imaging techniques; precision agriculture; precise fertilization

0 引言

据农业农村部2024年数据显示，我国每年肥料使用量约 49880kt 我国肥料利用率约占 35% ，远低于世界发达国家 50%～60% 的水平。（剩余15267字）