Rapid detection of corn moisture content based on improved ICEEMDAN algorithm combined with TCN-BiGRU model

Area of Science:

• Agricultural Science
• Spectroscopy
• Machine Learning

Background:

• Accurate corn moisture content (MC) is crucial for cultivation, harvesting, and storage.
• Traditional MC detection methods are slow, labor-intensive, and cumbersome.
• Developing rapid and efficient MC detection is essential for the corn industry.

Purpose of the Study:

• To develop a rapid and accurate corn MC detection method.
• To overcome the limitations of traditional MC detection techniques.
• To enhance corn yield, quality, and economic benefits through precise MC management.

Main Methods:

• Utilized Near-Infrared (NIR) spectral data from 405 corn seed samples.
• Applied improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN), optimized by the Crested Porcupine Optimizer (CPO), for noise reduction.
• Employed the Chaotic-Cuckoo Search (CCS) algorithm for characteristic wavenumber extraction.
• Developed a Temporal Convolutional Network-Bidirectional Gated Recurrent Unit (TCN-BiGRU) model for MC classification.

Main Results:

• The developed CPO-ICEEMDAN-CCS-TCN-BiGRU model achieved an accuracy of 97.54% for corn MC detection.
• The proposed model significantly outperformed existing methods like CNN, LSTM, TCN, PLS, and SVM.
• The accuracy improvement over other models ranged from 2.34% to 9.22%.

Conclusions:

• The CPO-ICEEMDAN-CCS-TCN-BiGRU model offers a reliable and efficient solution for rapid corn MC detection.
• This advanced method provides a strong foundation for optimizing corn production and management.
• The findings contribute to improving corn quality and maximizing economic returns.