New Publication: KBTrack Advances Automated Inventory Management for Ornamental Nurseries

The Smart Systems Engineering Lab is excited to announce the publication of our latest research on KBTrack, an AI-driven plant tracking and counting framework for ornamental nursery inventory management, in Computers and Electronics in Agriculture.

 

Accurate nursery inventory remains a major challenge because plants are densely arranged, frequently overlap in camera views, and can be difficult to track consistently over long video sequences. KBTrack addresses these challenges by combining deep learning-based plant detection and instance segmentation with a long-term identity association strategy designed to recover plant identities after occlusion. The framework is integrated into a cloud-enabled system that connects field-collected imagery with georeferenced nursery information for scalable inventory monitoring.

 

Field experiments demonstrated strong performance, with the ensemble model achieving a 0.982 mAP@50 for plant detection and 0.981 mAP@50 for instance segmentation. KBTrack achieved approximately 98.7% counting accuracy while substantially reducing identity-switching errors compared with conventional multi-object tracking approaches.

 

This work demonstrates how purpose-built computer vision and tracking methods can help replace labor-intensive manual inventory practices with more accurate, scalable, and data-driven approaches. The research contributes to the Smart Systems Engineering Lab's broader efforts in artificial intelligence, robotics, machine vision, and precision horticulture.

 

Congratulations to Mohtasim Hadi Rafi, Hamid Syed, Faraz Ahmad, Dr. Jeremy Pickens, and Dr. Tanzeel U. Rehman on this publication!