Team members: Mohamed Ali, Faranguisse Sadrieh, Kunj Golwala
The crab loading project addresses labor shortages and food processing safety concerns by developing an AI-driven, vision-guided robotic system for handling disordered agricultural products. Using Chesapeake Blue Crabs as a test case, the system combines imaging, computer vision, and robotics. A novel dual-laser scanning technique produces high-resolution depth maps, enabling accurate topographic modeling. Advanced vision algorithms, including transformer-based networks like Mask2Former, perform highly precise instance segmentation and pose estimation. A Universal Robot UR5e, equipped with a soft pneumatic gripper and integrated via ROS middleware, executes precise picking and placing tasks. Validated by Chesapeake blue crabs, this system significantly improves accuracy, speed, and operational efficiency.