Marker Intelligence: Redefining Vision-Based Tactile Sensor Performance through Structured Marker Design

Tactile perception is fundamental to manipulation and intelligent sensing, yet the impact of marker design in vision-based tactile sensors(VTS) remains largely unexplored. Conventional approaches rely on tracking randomly distributed markers with arbitrary shapes, assuming displacement alone suffices to characterize sensor deformation. However, this ignores the potential of structured marker configurations to improve sensor performance. We introduced Marker Intelligence, a new paradigm that transforms optical markers from passive tracking tools into active intelligent sensing components. By structuring markers to align with natural deformation patterns, Marker Intelligence improves force prediction and classification accuracy. We developed a framework integrating PCA, Hu moments, and RSS fitting to quantify marker performance. Our results showed that circular markers consistently outperformed other shapes, achieving higher accuracy with newly introduced concentric ring design, surpassing the Gelsight benchmark. These findings establish Marker Intelligence as a foundational principle in VTS field, advancing sensor optimization and perception.