Light adaptive image enhancement for improving visual analysis in intercropping cultivation.

Area of Science:

• Agricultural Science
• Computer Vision
• Remote Sensing

Background:

• Intercropping maize and soybean presents challenges for image-based recognition due to varying plant heights and resulting shadows.
• Illumination variation significantly impacts the accuracy of image analysis in agricultural settings.

Purpose of the Study:

• To develop and evaluate an illumination compensation model for soybean canopy images in maize-soybean intercropping systems.
• To enhance image uniformity and mitigate the effects of uneven lighting for improved crop recognition.

Main Methods:

• A regression model was developed using crop height differences, solar elevation angle, and light intensity.
• The model was applied to correct soybean canopy images and compared with histogram equalization, Multi-Scale Retinex (MSR), and gamma correction.
• Quantitative evaluation used peak signal-to-noise ratio (PSNR) and analysis of RGB and HLS color channels.

Main Results:

• The proposed illumination compensation model achieved a PSNR of 40.79 dB, indicating superior image quality.
• The method effectively increased brightness and reduced local fluctuations, with noticeable improvements in green channel values and overall RGB values.
• Channel-wise standard deviation analysis showed lower variance in green (G) and hue (H) channels, demonstrating improved image consistency.

Conclusions:

• The developed illumination compensation model significantly enhances image uniformity in maize-soybean intercropping systems.
• This advancement is crucial for improving the accuracy and reliability of image-based recognition tasks in complex agricultural environments.
• The model offers a robust solution for addressing illumination variations, paving the way for more effective precision agriculture applications.