An improved vision-based tactile skin with imaging adjustment system to reduce defocusing caused by contact depth changes

Abstract

Imaging robustness of the vision-based tactile sensor (VTS) depends on coating wear resistance and imaging stability. With the miniaturization of sensors, the limited imaging space cannot afford a buffer against imaging deviations caused by contact depth changes. This paper proposes an improved vision-based thin tactile sensor (VTTS, also named vision-based tactile skin), integrating a focus-adjustable camera and an imaging adjustment system (IAS) to improve imaging robustness. The IAS contains an imaging calibration system and a focusing system. Since visuo-tactile sensing is performed in a relatively stable and closed imaging environment, the image-based focusing method is naturally matched with tactile imaging. Accordingly, an imaging calibration system based on a global search is designed to determine the focusing interval and shorten the focusing range. Three-point-fitting-based depth from focus (TPF-DFF) and deep-learning-based depth from defocus (DL-DFD) are proposed for low-frequency (simple)/high-frequency (complex) imaging adjustment. The experimental results demonstrate that the IAS-VTTS has close to 100% adjustment accuracy, and the fastest focusing time is up to 40 ms. Consequently, the IAS-VTTS can adapt to different deformation depths and support high-quality tactile imaging.