人工智能深度学习模型在土壤属性数字制图中的应用
伍维模1
作者信息
- 1. 塔里木大学 农学院,新疆 阿拉尔 843300;塔里木大学 南疆绿洲农业资源与环境研究中心,新疆 阿拉尔 843300;南疆干旱区作物遗传改良与高效生产重点实验室,新疆 阿拉尔 843300
- 折叠
摘要
为提高土壤属性数字制图预测精度,以及随着遥感环境变量数据量的增加、算力的增强和开源深度学习框架的普及,数字土壤制图正在从传统的知识驱动模型向数据驱动的人工智能深度学习模型转变.该文以土壤关键属性有机碳为例,分析归纳土壤有机碳数字制图深度学习模型的理论基础、模型结构、亟待解决的有关环境变量空间上下文信息和多模态数据整合及模型可解释性等问题,旨在促进人工智能深度学习模型在第三次全国土壤普查土壤属性制图中的应用.
Abstract
In order to increase the prediction accuracy of digital mapping model of soil properties,accompanied by the increase in data quantity especially remote sensing environmental variables,the improvement of computing power,and the popularization of deep learning frameworks,the digital mapping model of soil propertiesis transitioning from traditional knowledge-driven models to data-driven artificial intelligent deep learning models.This article takes the key property of soil organic carbon as an example to analyze and summarize the theoretical basis,model structure,integration of relevant environmental variable spatial context information and multimodality data that urgently needs to be solved,and interpretability of deep learning models.The aim is to promote the application of artificial intelligent deep learning models in soil properties digital mapping of the Third National Soil Survey.
关键词
数字土壤制图/深度学习/神经网络/土壤属性/土壤有机碳Key words
digital soil mapping/deep learning/neural network/soil properties/soil organic carbon引用本文复制引用
基金项目
塔里木大学高教研究项目(TDGJYB2226)
出版年
2024
国家科技期刊平台
NSTL
万方数据