An Object Feature-Based Recognition and Localization Method for Wolfberry.

Area of Science:

• Agricultural Robotics
• Computer Vision
• Image Processing

Background:

• Object recognition and localization are critical for automated harvesting robots.
• Unstructured lighting and occlusion pose significant challenges in agricultural environments.

Purpose of the Study:

• To develop an advanced image segmentation algorithm for wolfberry harvesting robots.
• To enhance the accuracy of segmenting wolfberry fruits and branches under complex conditions.

Main Methods:

• A feature fusion algorithm combining Lab (a-channel) and YIQ (I-channel) color spaces with wavelet transformation for fruit segmentation.
• A K-means clustering algorithm in the Lab color space, coupled with morphological processing and length filtering for branch segmentation.
• Localization of gripping point coordinates for branches.

Main Results:

• Achieved 78% segmentation accuracy for wolfberry fruits in 500 samples under complex lighting and occlusion.
• Demonstrated high accuracy in precise branch segmentation and localization of gripping points.
• The algorithm effectively handles illumination changes and occlusion.

Conclusions:

• The proposed algorithm significantly improves segmentation and localization accuracy compared to traditional methods.
• Provides essential technical support for the vision systems of field-based wolfberry harvesting robots.
• Offers a practical reference for automated agricultural harvesting operations.