【摘要】：提出一種基于深度稀疏學習的土壤近紅外光譜分析預測模型。首先,使用稀疏特征學習方法對土壤近紅外光譜數(shù)據(jù)進行約簡,實現(xiàn)土壤近紅外光譜內(nèi)容的稀疏表示;然后采用徑向基函數(shù)神經(jīng)網(wǎng)絡以稀疏表示特征系數(shù)為輸入,以所測土壤成分為輸出,分別建立土壤有機質、速效磷、速效鉀的非線性預測模型。結果表明用該模型預測土壤有機質的含量是可行的,但對土壤速效磷和速效鉀含量的預測還需對模型做進一步的優(yōu)化。
[Abstract]:A soil near infrared spectroscopy (NIR) analysis and prediction model based on deep sparse learning is proposed. Firstly, the sparse feature learning method is used to reduce the soil near infrared spectral data to realize the sparse representation of the soil near infrared spectrum, and then the sparse representation feature coefficient is used as the input of the radial basis function neural network. The nonlinear prediction models of soil organic matter, available phosphorus and available potassium were established by using the measured soil composition as the output. The results showed that it was feasible to predict the content of soil organic matter by using this model, but the prediction of soil available phosphorus and potassium content needed to be further optimized.
【作者單位】： 中國科學院合肥智能機械研究所;
【基金】：中國科學院科技服務網(wǎng)絡計劃(KFJ-EW-STS-069) 國家自然科學基金(31671586)資助項目~~
【分類號】：S151.9;O657.33;TP183

【相似文獻】

相關期刊論文 前10條

1 李勇,魏益民,王鋒;影響近紅外光譜分析結果準確性的因素[J];核農(nóng)學報;2005年03期

2 褚小立,袁洪福,陸婉珍;基礎數(shù)據(jù)準確性對近紅外光譜分析結果的影響[J];光譜學與光譜分析;2005年06期

3 楊皓e，

本文編號：2151973