Vision feedback control method based on wavelet nerve network

Abstract

The invention is a control method for visual feedback based on a wavelet neural network, and relates to the control method for the visual feedback of a robot, and belongs to the field of artificial intelligence. The invention puts forward, designs and implements a control method for visual feedback based on a wavelet neural network aimed at the problem that a video camera model, a robot model and the hand-eye relationship need calibrating in the traditional control method for the visual feedback of a robot. In the invention, a camera (the eyes of a robot) is aimed at a target object in a workplace and an image is shot. An image characteristics extraction is conducted by a computer, fed back to the input end of the system and then compared with the expected value. If the two (the image characteristics and the expected value) are unequal, an expected key angle value of the robot is obtained through the wavelet neural network, and then the movement of the robot (camera) is controlled by a robot controller. The process is repeated until the difference of the image characteristics is zero. The method adopts a wavelet neural network, thereby, the demand for calibrating each part of the robot system is avoided. The method adopts the characteristic space to obtain the image characteristics through transformation and is applicable to different target objects.

Description

technical field [0001] The invention relates to a robot visual feedback control method and belongs to the field of artificial intelligence. Background technique [0002] In recent years, efforts have been made to make robots have human-like visual capabilities. The current common practice is to use various off-the-shelf imaging devices (such as CCD cameras). Obtain environmental information through imaging equipment, process the information and feed it back to the input end of the system to guide and coordinate the robot to perform various operations. Similar research has important practical value in fields such as industry, agriculture, service industry, national defense and aerospace. However, traditional methods require that the camera model, robot model, and hand-eye relationship are known or partially known, which requires a lot of time-consuming and laborious calibration work. In order to avoid or simplify the calibration process of the system, researchers have prop...

AI Technical Summary

Problems solved by technology

The advantage of this method is that no system calibration is required, but the disadvantage is that each iteration needs to find the inverse matrix of the Jacobian matrix

For complex robot systems, the image Jacobian matrix has a high dimension, so it is very difficult to calculate the Jacobian inverse matrix online. In addition, when the Jacobian matrix is ​​singular, the absence of the inverse matrix will cause control failure

Images