基于改进GPR的无人机采集田间水稻图像颜色校正方法

打开文本图片集

中图分类号：S511；TP391 文献标识码：A 文章编号：2095-5553（2026）04-0171-06

Abstract：Color corectionoffieldrice images capturedbyunmannedaerialvehicles（UAVs）is essential to ensure that the images accuratelyreflect theactual growth statusof ricecrops，therebysupporting beteragricultural decision-making and precision farming practices.Toaddressthe limitationof existing models in achieving efectivecolor corection，this paper proposed a novel method based on Enhanced Gaussan Process Regression （EGPR）.The EGPR approach improved upon traditionalGaussian processregresion byemployinga composite kernelfunctioninplaceofasinglekermel，thereby enhancing modeling flexibilityandacuracy.Toreduce thecomputational burden associatedwiththecomposite kernel function，the WhaleOptimization Algorithmwasusedtooptimizethe model'shyperparameters eficiently.Experimental validationinvolvedcapturing imagesofa24-colorreferencechartundervarying lightingconditionsand UAVexposure setings.Theresults demonstrated that UAV exposurecompensation significantly influencedthecolorqualityof the images. Notably，images taken between 8：O0 and 1O：OO a.m. with an exposure compensation of -1.3 produced the most accurateandstablecolor representation throughout the day.The EGPR effectivelyenhanced the color performance of UAV-acquired field images，maintainingacolor diference value within5.This markeda notableimprovement over traditionalcolorcorectionmodels.Thisresearchprovidesareliabletechnicalfoundationfortheacurateinterpretationand application of UAV-collected rice field imagery in agricultural monitoring and management.

Keywords：unmanned aerial vehicle;rice image;；color correction; Gaussian Process Regression; adaptive optimization

0 引言

近年来，我国的无人机遥感技术在多个领域都取得了显著进展并得到广泛应用，包括农业、环境监测、城市规划和资源管理等[1.2]。（剩余9397字）