Fast Detection of Tomato Sucker Using Semantic Segmentation Neural Networks Based on RGB-D Images

Area of Science:

• Agricultural Engineering
• Computer Vision
• Machine Learning

Background:

• Tomato sucker removal is crucial for plant yield and health.
• Manual removal is labor-intensive and time-consuming for farmers.
• Automated solutions are needed to optimize tomato plant care.

Purpose of the Study:

• To develop an automated system for detecting tomato suckers.
• To improve the efficiency of tomato plant management through technology.
• To reduce labor costs and increase crop yield in tomato farming.

Main Methods:

• Proposed a novel semantic segmentation neural network for tomato sucker detection.
• Utilized RGB-D images to capture both visual and depth information.
• Created a dedicated tomato RGB-D image dataset for training and evaluation.

Main Results:

• The proposed network achieves real-time performance at 138.2 frames per second.
• Demonstrated a sucker detection accuracy of 80.2%.
• The network is computationally efficient with only 680,760 parameters, requiring low system resources.

Conclusions:

• The developed semantic segmentation network offers an effective and efficient solution for automated tomato sucker detection.
• Its real-time capability and low resource requirement make it suitable for practical agricultural applications.
• This technology has the potential to significantly aid farmers in tomato cultivation.