Semi-automatic normalization of multitemporal remote images based on vegetative pseudo-invariant features.

Area of Science:

• Remote Sensing
• Agricultural Science
• Image Processing

Background:

• Multitemporal remote sensing images are crucial for monitoring agricultural landscapes.
• Variability in image acquisition conditions can introduce noise and hinder accurate analysis.
• Standardized image data is essential for reliable change detection and vegetation monitoring.

Purpose of the Study:

• To develop and validate a semi-automatic procedure for normalizing multitemporal remote sensing images of agricultural scenes.
• To reduce spectral variability and improve the comparability of images acquired at different times.
• To create a software tool (ARIN) for efficient image normalization.

Main Methods:

• The ARIN procedure identifies pseudo-invariant features (VPIFs) across multiple images.
• Spectral data from VPIFs are used to calculate correction factors (CFs).
• Linear transformation with CFs normalizes image bands, and ARIN software automates this process.

Main Results:

• ARIN significantly reduced the range, standard deviation, and RMSE of spectral bands and vegetation indices.
• Normalization efficacy was consistent across different VPIFs (citrus, olive, poplar).
• High correlation coefficients (≥0.85) between CFs across VPIFs confirmed normalization consistency.

Conclusions:

• The ARIN method provides effective semi-automatic relative image normalization for agricultural and forestry landscapes.
• The procedure ensures comparable results regardless of the selected VPIFs.
• ARIN enhances the reliability of multitemporal remote sensing data for agricultural monitoring.