Crop row detection in maize fields inspired on the human visual perception

Area of Science:

• Agricultural Engineering
• Computer Vision
• Robotics

Background:

• Mobile agricultural vehicles experience vibrations and movements affecting onboard vision systems.
• Accurate crop row detection is crucial for precision agriculture tasks like automated weeding and harvesting.
• Existing methods like Hough transformation can be sensitive to image distortions and noise.

Purpose of the Study:

• To develop a robust real-time image processing method for identifying maize crop rows.
• To address challenges posed by vehicle motion and image perspective distortions.
• To provide a reliable system for agricultural vehicle navigation and operation.

Main Methods:

• Image segmentation using fuzzy clustering to differentiate green pixels (crops/weeds) from background.
• Crop row detection via image perspective projection to find alignments of segmented green pixels.
• Real-time processing for onboard agricultural vehicle application.

Main Results:

• The proposed method successfully identifies crop rows in maize fields under challenging conditions.
• The fuzzy clustering approach effectively determines optimal thresholds for image segmentation.
• The perspective projection method robustly detects crop lines, outperforming Hough transformation in comparative tests.

Conclusions:

• The developed image processing technique offers a robust solution for real-time crop row detection in maize.
• The method's resilience to vibrations and perspective effects makes it suitable for mobile agricultural platforms.
• This technology enhances the potential for autonomous operations in precision agriculture.