YOLOF-Snake: An Efficient Segmentation Model for Green Object Fruit

Area of Science:

• Agricultural Engineering
• Computer Vision
• Machine Learning

Background:

• Accurate fruit detection and segmentation are crucial for orchard automation.
• Complex orchard environments with variable lighting and weather pose challenges for green fruit segmentation.

Purpose of the Study:

• To propose an efficient YOLOF-snake segmentation model for accurate green fruit detection and segmentation.
• To address the challenges of segmenting green fruits in complex orchard backgrounds.

Main Methods:

• Utilized ResNet101 as the backbone network for feature extraction.
• Employed a decoder with expanded receptive fields for classification and regression.
• Integrated a Deep-snake network with a diamond-shaped structure derived from regression box center points for contour adjustment.

Main Results:

• The YOLOF-snake model demonstrated high sensitivity to green fruits.
• Significant improvements in segmentation accuracy and efficiency were observed.
• The model effectively segments green fruits even in challenging conditions.

Conclusions:

• The YOLOF-snake model offers a robust solution for green fruit segmentation in orchards.
• This approach enhances the capabilities of agricultural equipment and provides a reference for other fruit and vegetable segmentation tasks.