On-the-cage robotic retrieval system using machine vision for digital management of quail egg production

• This study introduces a new concept for quail egg retrieval in poultry production directly at the cage level. • The proposed system consists of a 5-degree-of-freedom (5-DoF) cartesian robotic manipulator mounted on a conveyor belt integrated with a machine vision system. • R 3 Det, Oriented R-CNN, OBB RetinaNet, and YOLOv8x-OBB—were evaluated for performance in egg detection and orientation. • The theoretical picking rates ranged from 457 eggs/h to 290 and 149 eggs/h, corresponding to faster to slower operation cycles, respectively. • Overall, the proposed system supports the development of next-generation smart sensors for in-cage egg quality and sanitary monitoring, enabling more effective and traceable poultry production. In industrial quail farming settings, conventional egg retrieval systems rely on multiple conveyor belts connected to stacked layers of cages. Although they are effective, these systems lack precision in monitoring flocks and the quality of eggs at the level of individual cages. As a result, when defective eggs are laid, farmers are unable to trace their origin or understand the underlying causes, creating a gap in identifying sanitary risks and preventing the spread of diseases and infections. This study introduces a new concept for quail egg retrieval in poultry production. The proposed system consists of a 5-degree-of-freedom (5-DoF) cartesian robotic manipulator with a mounted conveyor belt integrated with a machine vision system based on oriented localization algorithms for precise egg collection directly at the cage level. Four base model algorithms—R 3 Det, Oriented R-CNN, OBB RetinaNet, and YOLOv8x-OBB—were evaluated for performance in egg detection and orientation. The machine vision system achieved the highest accuracy when embedded with Oriented R-CNN, followed by OBB RetinaNet, R 3 Det, and YOLOv8x-OBB, with RMSE values of 2.9, 3.87, 5.67, and 11.2, respectively, for the oriented picking of eggs. A comparative evaluation of three stepper motor resolutions revealed trade-offs between speed and precision when eggs were picked from cages. The theoretical picking rates ranged from 457 eggs/h to 290 and 149 eggs/h, corresponding to faster to slower operation cycles, respectively. Overall, the proposed system supports the development of next-generation smart sensors for in-cage egg quality and sanitary monitoring, enabling more effective and traceable poultry production.