[Research on maize multispectral image accurate segmentation and chlorophyll index estimation]

Area of Science:

• Agricultural remote sensing
• Plant physiology
• Image processing technology

Background:

• Rapid acquisition of maize growing information is crucial for effective field management.
• Traditional chlorophyll measurement methods can be destructive and time-consuming.
• Non-destructive techniques are needed for efficient crop monitoring.

Purpose of the Study:

• To develop and validate a non-destructive method for measuring maize chlorophyll content index.
• To utilize multi-spectral imaging and image processing for rapid field data acquisition.
• To establish a model for diagnosing chlorophyll content in maize.

Main Methods:

• Acquisition of multi-spectral canopy images (visible and near-infrared bands) using a 2-CCD system.
• Image processing techniques including adaptive smoothing, variable threshold segmentation, and region labeling for leaf segmentation.
• Calculation of image parameters (average gray values) and vegetation indices, followed by partial least squares regression (PLSR) modeling.

Main Results:

• Accurate segmentation of maize leaves from background with 95.59% accuracy.
• Development of a chlorophyll index detection model using PLSR.
• The model achieved a predicting R² of 0.5685, demonstrating feasibility for chlorophyll diagnosis.

Conclusions:

• The developed non-destructive method based on multi-spectral imaging is feasible for diagnosing maize chlorophyll content.
• This technique enables rapid and efficient acquisition of crucial crop growing information.
• The study provides a foundation for advanced precision agriculture applications in maize cultivation.