郭 倩, 王会利, 王晓晴, 杨华蕾, 张美薇, 曾令涛, 崔宇培, 徐富义, 孙孝林. 基于智能手机图像颜色参数的土壤有机质估测[J]. 土壤通报, 2024, 55(4): 932 − 943. DOI: 10.19336/j.cnki.trtb.2023120605
引用本文: 郭 倩, 王会利, 王晓晴, 杨华蕾, 张美薇, 曾令涛, 崔宇培, 徐富义, 孙孝林. 基于智能手机图像颜色参数的土壤有机质估测[J]. 土壤通报, 2024, 55(4): 932 − 943. DOI: 10.19336/j.cnki.trtb.2023120605
GUO Qian, WANG Hui-li, WANG Xiao-qing, YANG Hua-lei, ZHANG Mei-wei, ZENG Ling-tao, CUI Yu-pei, XU Fu-yi, SUN Xiao-lin. Estimation of Soil Organic Matter Based on Color Parameters of Smartphone Images[J]. Chinese Journal of Soil Science, 2024, 55(4): 932 − 943. DOI: 10.19336/j.cnki.trtb.2023120605
Citation: GUO Qian, WANG Hui-li, WANG Xiao-qing, YANG Hua-lei, ZHANG Mei-wei, ZENG Ling-tao, CUI Yu-pei, XU Fu-yi, SUN Xiao-lin. Estimation of Soil Organic Matter Based on Color Parameters of Smartphone Images[J]. Chinese Journal of Soil Science, 2024, 55(4): 932 − 943. DOI: 10.19336/j.cnki.trtb.2023120605

基于智能手机图像颜色参数的土壤有机质估测

Estimation of Soil Organic Matter Based on Color Parameters of Smartphone Images

  • 摘要:
    目的 应用深度学习模型开展基于智能手机图像颜色参数的土壤有机质估测,并评价应用效果。
    方法 在自制光学暗室中拍摄728个风干、过筛后的土壤样本,对所得图像进行预处理后得到其红色(R)、绿色(G)、蓝色(B)、色调(H)、饱和度(S)和明度(V)六个颜色通道的中值和均值,分别建立卷积神经网络(CNN)、长短期记忆网络(LSTM)和随机森林(RF)模型,并使用基于树形结构Parzen估计器的贝叶斯方法进行参数优化,进而估测土壤有机质(SOM)含量,并对估测结果的准确度进行五折交叉验证。
    结果 CNN、LSTM和RF三种模型均有着良好的估测准确度,决定系数(R2)范围为0.732 ~ 0.856,均方根误差(RMSE)范围为4.721 ~ 6.455 g kg−1,一致性相关系数(CCC)范围为0.843 ~ 0.917;其中,基于全部颜色参数的RF模型有着最优的估测效果,且R值、V值是该模型中最重要的参数;总体上,估测准确度的排序为RF略优于CNN,而CNN又略优于LSTM。
    结论 CNN和LSTM在基于智能手机图像颜色参数估测土壤有机质中有着良好的准确度,但在小样本量的情况下,CNN、LSTM模型的准确度略低于RF。

     

    Abstract:
    Objective Accurate estimation of soil organic matter (SOM) content has always attracted a lot of attention. Compared with traditional laboratory analysis methods and professional spectrometers, SOM estimation based on color parameters of smartphone images shows advantages in more economic, higher efficiency and greater convenience. The aim of this study was to assess performance of deep learning on estimating SOM based on color parameters of smartphone images.
    Method 728 dry soil samples were photographed in a self-made optical darkroom and determined for SOM. Then, median and mean values of six color channels, i.e., Red (R), Green (G), Blue (B), Hue (H), Saturation (S), Value (V) were obtained after gray processing, threshold segmentation, color conversion and other preprocessing, based on the obtained image through Python and MATLAB software. Subsequently, convolutional neural network (CNN), long short-term memory network (LSTM) and random forest (RF) models were established for SOM estimation, respectively,. Accuracies of the models were evaluated through five-fold cross-validation.
    Result The results indicates that CNN, LSTM and RF models had excellent performances with coefficient of determination (R2) rangs from 0.732 to 0.856, the root mean square error (RMSE) ranging from 4.721 to 6.455, and the Lin's concordance correlation coefficient (CCC) ranging from 0.843 to 0.917. The comprehensive performance order was: RF model, CNN model, LSTM. R value and V value were the most important color parameters.
    Conclusion CNN and LSTM models demonstrated good accuracy in estimating SOM using color parameters from smartphone images. However, in the case of small sample sizes, the accuracy of CNN and LSTM models was slightly lower than that of RF.

     

/

返回文章
返回