研究目的
Investigating the effectiveness of a 3-D surface reconstruction method for optical tactile sensors with shadow detection and compensation for robotic fingers.
研究成果
The proposed optical tactile sensor with 3-D surface reconstruction and shadow detection has been validated through simulations and experiments. It effectively reconstructs the surface shape of objects with modest surface gradients, though it faces challenges with objects having very steep or complicated surfaces. Future work will focus on improving the sensing layer's physical properties and miniaturizing the sensor for integration into robotic fingers.
研究不足
The tactile sensor may introduce errors for object surface regions with very steep gradient changes. The sensing layer's softness is not sufficient to perceive small details of complicated object surfaces, leading to large reconstruction errors in such cases.
1：Experimental Design and Method Selection:
The tactile sensor hardware consists of a surface deformation sensing layer, an image sensor, and four individually controlled flashing LEDs. The method involves taking four deformation images of an object with different illumination directions, building look-up tables to map intensity distribution to image gradient data, detecting and amending image shadows, and reconstructing the 3-D depth distribution from the 2-D gradient.
2：Sample Selection and Data Sources:
A rigid sphere with a known radius is used as a calibration object to generate the reflectance map. Objects with varying surface textures are used for testing.
3：List of Experimental Equipment and Materials:
The prototype includes an elastic sensing layer made of soft silicone, an image sensor (Nikon D7000 camera with a Tamron lens), four white LEDs, and a thin metal film covering the sensing layer.
4：Experimental Procedures and Operational Workflow:
The sensor takes four images of an object with different LED illuminations. The images are processed to detect shadows, and the 3-D surface is reconstructed using the gradient data from the look-up tables.
5：Data Analysis Methods:
The gradient field is obtained from the intensity data using k-NN search and k-d tree for acceleration. The depth information is calculated from the gradient field using the Frankot–Chellappa algorithm.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
