Method for Identifying Maize Haploid Seeds by Applying Diffuse Transmission Near-Infrared Spectroscopy

Area of Science:

• Agricultural Science
• Spectroscopy
• Biotechnology

Background:

• Maize haploid breeding relies on accurate haploid seed identification.
• Near-infrared (NIR) spectroscopy with diffuse transmission (DT) enables automated selection.
• NIR spectra often contain redundant features and noise, hindering identification performance.

Purpose of the Study:

• To develop an improved method for automatic maize haploid seed identification.
• To address challenges of redundant features and noise in NIR spectra.
• To enhance the accuracy and efficiency of haploid seed selection in maize.

Main Methods:

• Utilized zero-phase component analysis (ZCA) for spectral feature space uniformity.
• Employed partial least squares regression (PLSR) for feature space dimensionality reduction.
• Developed a back propagation neural network (BPNN) classifier for identification.

Main Results:

• Achieved an outstanding average identification accuracy of 96.16%.
• Demonstrated a minor standard deviation (SD) in accuracy, indicating method robustness.
• The proposed method significantly improved upon existing techniques for maize haploid seed identification.

Conclusions:

• The developed method offers a robust and highly accurate approach for automatic maize haploid seed identification.
• Feature space optimization using ZCA and PLSR effectively mitigates noise and redundancy.
• This technique holds significant potential for advancing maize haploid breeding programs.