A Soft Finger Gripper with Force Feedback Function Based on Image Processing and Neural Network

Soft fingers and the perception of force enable humans to grasp objects in a gentle and stable way. However, existing grippers do not have a good sense of the force. This paper aims to present the structure of a flexible mechanical gripper and the method of processing transformed images into dynamic data. First, we developed a mechanical gripper that consists of three soft Fin-Ray fingers with three cameras attached. Four parallel red stripes are attached to the surface of the soft finger, and the deformation occurs when the surface of the soft Fin-Ray finger is compressed by the pinched objects. When different amounts of force are applied, the images of the paralleled red lines taken by the camera show their displacements. By processing the image, the exact position of the red stripes in the image can be obtained. The position of the red stripes in the image is associated with the force at this time by BP neural network. After the training of the neural network, the force recognition accuracy reaches 92.5%.