Crop Classification Based on Red Edge Features Analysis of GF-6 WFV Data

Area of Science:

• Agricultural remote sensing
• Spectral analysis
• Geospatial data analysis

Background:

• Crop classification accuracy is crucial for agricultural monitoring.
• Red edge spectral bands offer unique insights into vegetation health and type.
• GF-6 WFV satellite data provides multiple red edge bands for detailed analysis.

Purpose of the Study:

• To analyze the influence of different red edge features on crop classification accuracy using GF-6 WFV data.
• To evaluate the effectiveness of red edge spectral, texture, and index features for crop identification.
• To compare feature selection and importance evaluation methods for remote sensing data.

Main Methods:

• Utilized GF-6 WFV satellite data from Hengshui City, China.
• Performed spectral analysis on red edge bands (710 nm and 750 nm).
• Applied stepwise discriminant analysis (SDA) and random forest (RF) for feature selection and classification.

Main Results:

• The red edge 710 band provided better crop separability and classification accuracy than the red edge 750 band.
• Random Forest (RF) outperformed SDA in feature importance evaluation.
• Red edge spectral features, texture features, and indices improved classification accuracy, with red edge 710 band features being most effective.

Conclusions:

• Red edge bands, particularly at 710 nm, are valuable for enhancing crop classification accuracy in remote sensing.
• RF is a superior method for feature importance evaluation in this context.
• The findings support the application of GF-6 WFV data and its red edge bands for improved agricultural monitoring.