Using a Region-Based Convolutional Neural Network (R-CNN) for Potato Segmentation in a Sorting Process.

Area of Science:

• Agricultural Engineering
• Computer Vision
• Machine Learning

Background:

• Automated sorting systems require efficient image segmentation for high-throughput processing.
• Traditional methods may lack the precision needed for complex agricultural sorting tasks.
• Developing robust segmentation is crucial for improving the performance of potato-sorting machinery.

Purpose of the Study:

• To implement and evaluate Mask R-CNN for precise potato segmentation.
• To compare Mask R-CNN's segmentation accuracy against classic Convolutional Neural Network (CNN)-based object detectors.
• To optimize Mask R-CNN parameters for superior segmentation results in potato sorting.

Main Methods:

• Utilized a region-based Convolutional Neural Network (R-CNN) approach, specifically Mask R-CNN.
• Focused on the segmentation component of a camera-input-based potato sorting and classification system.
• Evaluated Mask R-CNN performance across various parameter settings to determine optimal configurations.

Main Results:

• Mask R-CNN demonstrated effective and precise segmentation of potatoes.
• The implemented Mask R-CNN models showed potential for integration into production environments.
• Achieved improved segmentation accuracy compared to traditional CNN-based object detection methods.

Conclusions:

• Mask R-CNN is a viable and effective tool for potato segmentation in automated sorting.
• The findings support the use of Mask R-CNN to enhance the efficiency and accuracy of potato sorting processes.
• This research contributes to the advancement of intelligent agricultural automation systems.