Smartphone-Based SPAD Value Estimation for Jujube Leaves Using Machine Learning: A Study on RGB Feature Extraction and Hybrid Modeling.

Journal: Sensors (Basel, Switzerland)
PMID:

Abstract

Chlorophyll content in date leaves is critical for fruit quality and yield. Traditional detection methods are usually complex and expensive. This study proposes a rapid detection method for chlorophyll content using smartphone images and machine learning and deep learning models. The SPAD values and RGB images of Xinjiang date palm were collected. The RGB images were preprocessed and their color features were extracted using Python and OpenCV. Through correlation analysis, 21 color features highly correlated with chlorophyll content were selected and downscaled with principal component analysis. Models including SVR, RVM, CNN, CNN-SVR, and CNN-RVM were used for prediction. Among them, the CNN-SVR model showed the most stable performance with R values of 72.21% and 77.44% on the training and validation sets, respectively, which outperformed the other models. The proposed method is simple, cost-effective, and highly accurate, providing a novel technical approach for accurate management and health monitoring in the date industry. This method has the potential for wide application.

Authors

  • Qi Wang
    Biotherapeutics Discovery Research Center, Shanghai Institute of Materia Medica, Chinese Academy of Sciences, Shanghai, 201203, China.
  • Ziyan Shi
    College of Information Engineering, Tarim University, Alaer 843300, China.
  • Kaiyao Hou
    College of Information Engineering, Tarim University, Alaer 843300, China.
  • Ning Yan
    College of Information Engineering, Tarim University, Alaer 843300, China.
  • Cuiyun Wu
    The Third People's Hospital of HeFei, Heifei 230000, China.
  • Xu Li
    Department of Critical Care Medicine, The First Affiliated Hospital of China Medical University, Shenyang, Liaoning, China.