基于手机相片和机器学习算法的滨海土壤盐分含量预测  

Prediction of Coastal Soil Salinity Content Based on Mobile Phone Photos and Machine Learning Algorithms

作　　者：钮一可 周宇菲 侍晓颖 王婉婷 许新怡 徐璐 NIU Yike;ZHOU Yufei;SHI Xiaoying;WANG Wanting;XU Xinyi;XU Lu(School of Geography,Geomatics and Planning,Jiangsu Normal University,Xuzhou,Jiangsu 221116)

机构地区：[1]江苏师范大学地理测绘与城乡规划学院,江苏徐州221116

出　　处：《热带农业工程》2025年第1期23-28,共6页Tropical Agricultural Engineering

基　　金：国家自然科学基金项目(No.42371053);大学生创新创业训练计划项目(No.202210320103Y);大学生创新创业训练计划项目(No.XSJCX13011)。

摘　　要：选择江苏省盐城市沿海地区的裸露土壤为研究对象,于多变天气条件下采集土壤样本,同步获取相应土壤区域的智能手机图像,通过室内测试获得土壤电导率(EC),利用RStudio软件处理采集的土壤图像。结果显示,土壤电导率与图像颜色色相参数间存在显著相关性,色相与亮度、亮度与纯度间存在相关性;采用随机森林模型建模70%土壤数据,运用留一交叉验证法(LOOCV)验证模型,剩余30%数据用于模型测试,该过程重复100次以获得较高精度模型(R_(val)^(2)=0.94,RMSE_(val)=1.92,RPD_(val)=4.30)。通过模型参数重要性分析发现,颜色色相为预测土壤盐渍化的关键参数,其次为纯度和亮度。This study focused on exposed soils in the coastal areas of Yancheng City,Jiangsu Province,collecting soil samples under various weather conditions.Concurrently,smartphone images of the corresponding soil areas were captured.Soil electrical conductivity(EC)was measured indoors,and the collected soil images were processed using RStudio software.The results indicated a significant correlation between soil EC and the hue parameter in the image colors,as well as correlations between hue and brightness,and brightness and saturation.A random forest model was trained using 70%of the soil data,validated with leave-one-out cross-validation(LOOCV),and tested on the remaining 30%of the data.This process was repeated 100 times to achieve a high-precision model(R_(val)^(2)=0.94,RMSE_(val)=1.92,RPD_(val)=4.30).Analysis of the model's parameter importance revealed that hue is the key predictor for soil salinization,followed by saturation and brightness.

关 键 词：土壤盐渍化 手机相片 颜色空间 随机森林 

分 类 号：P232[天文地球—摄影测量与遥感] S152[天文地球—测绘科学与技术]