Research on GelSight materials and reconstruction algorithm based on visual touch
Peixin Niu, Jingjin Shen · 2024
Robot sensors play a crucial role in robotics. Through sensors, robots can sense and acquire environmental information to make intelligent decisions and act accordingly. Tactile sensors are particularly important because they enable robots to simulate human tactile perception and achieve high precision perception of surface details. As an innovative tactile sensor, GelSight sensor has the ability to capture and analyze the microstructure and deformation information of the surface of objects in real time, and has broad prospects for development. This high-resolution tactile perception provides the robot with a more accurate and comprehensive understanding of the environment, allowing the robot to interact with the environment more intelligently. However, the cost of GelSight sensor research is still relatively high, which limits its popularization in research and applications. To solve this problem, in this paper, we propose a low-cost version design scheme, and the use of carbonized particles to make the reflector. At the same time, the traditional photometric stereo algorithm is combined with the designed GelSight sensor hardware to explore the development prospect of this idea. The experimental results show that the low cost GelSight sensor version design and the new height reconstruction method are feasible and effective. These research results provide a new idea and method for further exploration and application of GelSight sensing technology, and have important theoretical significance and practical value for promoting the development of mechanical science and technology.