Graphic Visualization and Recognition System Based on Electroluminescent Devices and Robotic Arm

• • Recognition accuracy of 96.7% was achieved across seven shape classes (rectangle, triangle, circle, star, butterfly, flower, snowflake) using a CNN with an ESP32-CAM module, demonstrating reliable performance for robotic arm control in material-diverse object sorting tasks. • • The ACEL device employs a pyramidal conical structure luminescent layer, which enhances light extraction and spectral homogeneity, enabling multispectral imaging that reduces environmental light interference—critical for industrial robotics operating under variable illumination. • • The system integrates a flexible, low-power ACEL light source with a CNN, achieving high brightness and contrast while maintaining low power consumption, which extends operational lifetime and suits continuous robotic operation in smart manufacturing. • • Training convergence was validated through accuracy and loss curves over multiple rounds, with confusion matrix analysis confirming class-wise precision; this indicates the model's robustness for real-time deployment in dynamic environments, though further validation under extreme conditions is warranted.